commit acfd3caf591a9178bdb50da8fb35a46855e703a7
Author: John Smith <you@example.com>
Date:   Tue Oct 6 23:35:53 2026 -0400

    feat: implement Opencode Unlocked plugin for OpenClaude with harness emulation and protocol translation

diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
new file mode 100644
index 0000000..db186c6
--- /dev/null
+++ b/.claude-plugin/marketplace.json
@@ -0,0 +1,17 @@
+{
+  "name": "opencode-unlocked",
+  "owner": {
+    "name": "Opencode Unlocked Contributors",
+    "url": "https://github.com/opencode-unlocked"
+  },
+  "plugins": [
+    {
+      "name": "opencode-unlocked",
+      "source": "./",
+      "description": "Seamless OpenClaude provider plugin and authentic harness bridge for OpenCode Go models (default: Muse Spark 1.3 Contributor, Space Bunny, and more).",
+      "version": "1.0.0",
+      "category": "ai-providers",
+      "tags": ["opencode", "models", "provider", "muse-spark", "space-bunny"]
+    }
+  ]
+}
diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
new file mode 100644
index 0000000..7df3e10
--- /dev/null
+++ b/.claude-plugin/plugin.json
@@ -0,0 +1,17 @@
+{
+  "name": "opencode-unlocked",
+  "version": "1.0.0",
+  "description": "Seamless OpenClaude provider plugin and authentic harness bridge for OpenCode Go models.",
+  "author": {
+    "name": "Opencode Unlocked Contributors"
+  },
+  "license": "MIT",
+  "keywords": [
+    "opencode",
+    "opencode-go",
+    "muse-spark",
+    "space-bunny",
+    "provider",
+    "bridge"
+  ]
+}
diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..0611a5c
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,23 @@
+# Environment & Secrets (NEVER COMMIT)
+.env
+.env.*
+*.key
+*.token
+credentials.json
+
+# Local review repositories
+review/
+
+# Dependencies
+node_modules/
+package-lock.json
+
+# Runtime & Process State
+*.log
+*.pid
+.openclaude/
+.cache/
+
+# OS
+.DS_Store
+Thumbs.db
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..9e01e2f
--- /dev/null
+++ b/README.md
@@ -0,0 +1,148 @@
+# Opencode Unlocked (OpenClaude Plugin)
+
+A standalone plugin and native provider bridge for [OpenClaude](https://github.com/anthropics/openclaude) that unlocks full access to **OpenCode Go** subscription models (`https://opencode.ai/zen/go/v1`).
+
+Features:
+- **Authentic OpenCode v2 Harness Reverse Engineering**: Generates authentic 30-character descending session identifiers (`ses_<timestamp_base62>`), client affinity tracking, and full harness request headers (`User-Agent`, `x-opencode-client`, `x-opencode-session`, `x-opencode-project`, `x-session-affinity`).
+- **Seamless Protocol Translation**: OpenClaude talks standard OpenAI `/v1/chat/completions`. Upstream models like **`muse-spark-1.3-contributor`** require OpenAI `/responses` API. The bridge transparently translates requests and bi-directionally streams Server-Sent Events (SSE).
+- **Target Default Model**: Defaults to **`muse-spark-1.3-contributor`** with verified support for **`space-bunny`** and all 36+ OpenCode Go models.
+- **Model Name Normalizer**: Strips any client prefixes (`opencode-unlocked-*`, `opencode-go-*`) to prevent upstream rejection.
+- **Zero-Secret Distribution**: The repository contains no hardcoded keys or `.env` files. Users provide their key via environment variable, interactive command, or local config.
+- **Marketplace Ready**: Fully compatible with OpenClaude's plugin and marketplace system (`marketplace.json` + `plugin.json`).
+
+---
+
+## Quick Start
+
+### 1. Installation
+
+#### Option A: OpenClaude Marketplace
+In OpenClaude, add the repository marketplace and install the plugin:
+```bash
+/plugin marketplace add /path/to/opencode-4-openclaude
+/plugin install opencode-unlocked@opencode-unlocked
+```
+
+#### Option B: Direct Plugin Flag
+Launch OpenClaude directly with the plugin:
+```bash
+openclaude --plugin-dir /path/to/opencode-4-openclaude
+```
+
+---
+
+### 2. Configure Your API Key
+
+Provide your OpenCode Go API key (`oc_sk_...`) using any of the following methods:
+
+1. **CLI Command**:
+   ```bash
+   ./bin/opencode-unlocked set-key oc_sk_your_key_here
+   ```
+2. **Environment Variable**:
+   ```bash
+   export OPENCODE_UNLOCKED_API_KEY="oc_sk_your_key_here"
+   # or
+   export GO_KEY="oc_sk_your_key_here"
+   ```
+
+---
+
+### 3. Provider Setup for OpenClaude
+
+Run the setup utility to configure OpenClaude's profile:
+```bash
+node bin/setup.mjs
+```
+This automatically backs up your existing profile and sets:
+- `OPENAI_BASE_URL`: `http://127.0.0.1:8788/v1`
+- `OPENAI_MODEL`: `muse-spark-1.3-contributor`
+
+To restore your previous profile at any time:
+```bash
+node bin/setup.mjs restore
+```
+
+---
+
+## Daemon & CLI Management
+
+Manage the background bridge daemon using `./bin/opencode-unlocked`:
+
+```bash
+# Check status, active model, and token consumption
+./bin/opencode-unlocked status
+
+# Start daemon
+./bin/opencode-unlocked start
+
+# Stop daemon
+./bin/opencode-unlocked stop
+
+# Restart daemon
+./bin/opencode-unlocked restart
+
+# List all 36+ available models
+./bin/opencode-unlocked models
+
+# Switch active default model
+./bin/opencode-unlocked switch space-bunny
+
+# Run live connectivity verification
+./bin/opencode-unlocked test
+```
+
+---
+
+## Supported Models
+
+All 36+ OpenCode Go models are available, including:
+
+| Model ID | Protocol | Specialization |
+| :--- | :--- | :--- |
+| **`muse-spark-1.3-contributor`** | `/responses` (Translated) | **Target Default**: Deep Reasoning & Coding |
+| **`space-bunny`** | `/chat/completions` (Native) | Fast General & Code |
+| **`gpt-5.6-luna`** | `/responses` (Translated) | High-Capability Reasoning |
+| **`deepseek-v4-flash`** | `/chat/completions` (Native) | High-Throughput Flash |
+| **`glm-5.3`** | `/chat/completions` (Native) | General Language |
+| **`kimi-k2.5`** | `/chat/completions` (Native) | Long Context |
+| **`minimax-01`** | `/chat/completions` (Native) | Conversational |
+| **`qwen3.5-max`** | `/chat/completions` (Native) | General Capabilities |
+
+---
+
+## Architecture
+
+```
+OpenClaude CLI
+      │ (OpenAI /v1/chat/completions)
+      ▼
+Opencode Unlocked Gateway (127.0.0.1:8788)
+ ├── Harness Fingerprint (ses_ + descending base62 session ID)
+ ├── Session Affinity Cache (prompt caching preservation)
+ ├── Protocol Translation Engine
+ │     ├── muse-spark-1.3-contributor -> POST /responses
+ │     └── space-bunny / others       -> POST /chat/completions
+ └── Dynamic Model Catalog & Prefix Normalizer
+      │
+      ▼
+OpenCode Go Cloudflare Edge (https://opencode.ai/zen/go/v1)
+```
+
+---
+
+## Testing
+
+Run the test suite:
+```bash
+# Unit & integration tests
+npm test
+
+# Live model verification (requires API key)
+npm run test:live
+```
+
+---
+
+## License
+MIT
diff --git a/bin/opencode-unlocked b/bin/opencode-unlocked
new file mode 100755
index 0000000..e3116e9
--- /dev/null
+++ b/bin/opencode-unlocked
@@ -0,0 +1,222 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import path from "node:path";
+import { spawn } from "node:child_process";
+import { fileURLToPath } from "node:url";
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = path.dirname(__filename);
+const ROOT_DIR = path.resolve(__dirname, "..");
+const PID_FILE = path.join(ROOT_DIR, ".gateway.pid");
+const LOG_FILE = "/tmp/opencode-unlocked.log";
+
+const { CONFIG, getApiKey, saveUserConfig, loadUserConfig, getUserConfigPath } = await import("../server/config.mjs");
+const { catalog } = await import("../server/catalog.mjs");
+
+const args = process.argv.slice(2);
+const command = args[0] || "status";
+
+async function isGatewayRunning() {
+  try {
+    const res = await fetch(`http://${CONFIG.HOST}:${CONFIG.PORT}/health`, { signal: AbortSignal.timeout(1000) });
+    return res.ok;
+  } catch {
+    return false;
+  }
+}
+
+function getStoredPid() {
+  if (fs.existsSync(PID_FILE)) {
+    try {
+      const pid = parseInt(fs.readFileSync(PID_FILE, "utf-8").trim(), 10);
+      return isNaN(pid) ? null : pid;
+    } catch {
+      return null;
+    }
+  }
+  return null;
+}
+
+async function startDaemon() {
+  if (await isGatewayRunning()) {
+    console.log(`✔ Opencode Unlocked gateway is already running on http://${CONFIG.HOST}:${CONFIG.PORT}`);
+    return;
+  }
+
+  const out = fs.openSync(LOG_FILE, "a");
+  const err = fs.openSync(LOG_FILE, "a");
+
+  const child = spawn(process.execPath, [path.join(ROOT_DIR, "server", "gateway.mjs")], {
+    detached: true,
+    stdio: ["ignore", out, err],
+    cwd: ROOT_DIR,
+    env: process.env
+  });
+
+  child.unref();
+  fs.writeFileSync(PID_FILE, String(child.pid), "utf-8");
+
+  // Wait for health
+  for (let i = 0; i < 30; i++) {
+    await new Promise((r) => setTimeout(r, 100));
+    if (await isGatewayRunning()) {
+      console.log(`✔ Opencode Unlocked gateway started (PID: ${child.pid}) on http://${CONFIG.HOST}:${CONFIG.PORT}`);
+      console.log(`  Default Model: ${CONFIG.DEFAULT_MODEL}`);
+      console.log(`  Logs: ${LOG_FILE}`);
+      return;
+    }
+  }
+
+  console.error("❌ Failed to verify gateway startup. Check logs at:", LOG_FILE);
+  process.exit(1);
+}
+
+async function stopDaemon() {
+  const pid = getStoredPid();
+  if (pid) {
+    try {
+      process.kill(pid, "SIGTERM");
+      console.log(`✔ Sent stop signal to process ${pid}`);
+    } catch (e) {
+      if (e.code !== "ESRCH") console.error("Error stopping process:", e.message);
+    }
+    if (fs.existsSync(PID_FILE)) fs.unlinkSync(PID_FILE);
+  }
+
+  if (await isGatewayRunning()) {
+    console.log("Waiting for gateway to shut down...");
+    for (let i = 0; i < 20; i++) {
+      await new Promise((r) => setTimeout(r, 100));
+      if (!(await isGatewayRunning())) break;
+    }
+  }
+
+  if (!(await isGatewayRunning())) {
+    console.log("✔ Gateway stopped successfully.");
+  } else {
+    console.warn("⚠ Gateway is still responding on the port.");
+  }
+}
+
+async function printStatus() {
+  const running = await isGatewayRunning();
+  const apiKey = getApiKey();
+  const config = loadUserConfig();
+
+  console.log("==========================================");
+  console.log("        Opencode Unlocked Status          ");
+  console.log("==========================================");
+  console.log(`Gateway Status : ${running ? "🟢 ONLINE" : "🔴 OFFLINE"}`);
+  console.log(`Gateway URL    : http://${CONFIG.HOST}:${CONFIG.PORT}/v1`);
+  console.log(`Default Model  : ${config.defaultModel || CONFIG.DEFAULT_MODEL}`);
+  console.log(`API Key Loaded : ${apiKey ? `✔ Configured (${apiKey.slice(0, 8)}...)` : "❌ Missing"}`);
+  console.log(`Config Path    : ${getUserConfigPath()}`);
+
+  if (running) {
+    try {
+      const res = await fetch(`http://${CONFIG.HOST}:${CONFIG.PORT}/status`);
+      if (res.ok) {
+        const data = await res.json();
+        console.log("------------------------------------------");
+        console.log(`Session Req's  : ${data.stats.sessionRequests}`);
+        console.log(`Tokens Consumed: ${data.stats.sessionTokens.total} (in: ${data.stats.sessionTokens.input}, out: ${data.stats.sessionTokens.output})`);
+      }
+    } catch {
+      // ignore
+    }
+  }
+  console.log("==========================================");
+}
+
+async function listModels() {
+  console.log("Fetching available models...");
+  const models = await catalog.getModels(true);
+  console.log(`\nAvailable Models (${models.length}):`);
+  for (const m of models) {
+    const isDefault = m.id === CONFIG.DEFAULT_MODEL;
+    console.log(`  ${isDefault ? "★ [DEFAULT] " : "  - "}${m.id}`);
+  }
+}
+
+async function setKey(key) {
+  if (!key) {
+    console.error("Usage: opencode-unlocked set-key <OPENCODE_API_KEY>");
+    process.exit(1);
+  }
+  saveUserConfig({ apiKey: key });
+  console.log("✔ API key saved securely to:", getUserConfigPath());
+}
+
+async function setModel(modelName) {
+  if (!modelName) {
+    console.error("Usage: opencode-unlocked switch <MODEL_NAME>");
+    process.exit(1);
+  }
+  saveUserConfig({ defaultModel: modelName });
+  console.log(`✔ Default model updated to: ${modelName}`);
+}
+
+async function runTests() {
+  console.log("Running live verification suite...");
+  const child = spawn(process.execPath, [path.join(ROOT_DIR, "test", "live-models.test.mjs")], {
+    stdio: "inherit",
+    cwd: ROOT_DIR
+  });
+  child.on("exit", (code) => {
+    process.exit(code || 0);
+  });
+}
+
+switch (command) {
+  case "start":
+    await startDaemon();
+    break;
+  case "stop":
+    await stopDaemon();
+    break;
+  case "restart":
+    await stopDaemon();
+    await startDaemon();
+    break;
+  case "status":
+    await printStatus();
+    break;
+  case "models":
+    await listModels();
+    break;
+  case "set-key":
+  case "key":
+    await setKey(args[1]);
+    break;
+  case "switch":
+  case "set-model":
+    await setModel(args[1]);
+    break;
+  case "test":
+    await runTests();
+    break;
+  case "setup":
+    await import("./setup.mjs");
+    break;
+  case "help":
+  case "--help":
+  default:
+    console.log(`
+Opencode Unlocked CLI Management
+
+Usage: opencode-unlocked <command> [arguments]
+
+Commands:
+  start           Start the local bridge daemon
+  stop            Stop the running daemon
+  restart         Restart the daemon
+  status          Display daemon health, active model, and token usage
+  models          List all 36+ available OpenCode Go models
+  switch <model>  Change default model (e.g. space-bunny, muse-spark-1.3-contributor)
+  set-key <key>   Save OpenCode API key securely to user configuration
+  test            Run live verification on muse-spark and space-bunny
+  setup           Configure OpenClaude provider profile to use this gateway
+  help            Show this help dialog
+`);
+    break;
+}
diff --git a/bin/setup.mjs b/bin/setup.mjs
new file mode 100755
index 0000000..7bd1e02
--- /dev/null
+++ b/bin/setup.mjs
@@ -0,0 +1,124 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const homeDir = os.homedir();
+const openclaudeDir = path.join(homeDir, ".openclaude");
+const profilePath = path.join(openclaudeDir, ".openclaude-profile.json");
+const profileBackupPath = path.join(openclaudeDir, ".openclaude-profile.json.backup");
+
+const globalConfigPath = path.join(homeDir, ".openclaude.json");
+const globalConfigBackupPath = path.join(homeDir, ".openclaude.json.backup");
+
+const { CONFIG, loadUserConfig } = await import("../server/config.mjs");
+
+const PROVIDER_ID = "provider_opencode_unlocked";
+const PROVIDER_NAME = "Opencode Unlocked";
+
+export function configureOpenClaudeProfile(options = {}) {
+  const userConfig = loadUserConfig();
+  const targetModel = options.model || userConfig.defaultModel || CONFIG.DEFAULT_MODEL;
+  const targetPort = options.port || CONFIG.PORT;
+  const targetHost = options.host || CONFIG.HOST;
+  const baseUrl = `http://${targetHost}:${targetPort}/v1`;
+
+  // 1. Configure plural provider profile in ~/.openclaude.json
+  if (fs.existsSync(globalConfigPath)) {
+    if (!fs.existsSync(globalConfigBackupPath)) {
+      fs.copyFileSync(globalConfigPath, globalConfigBackupPath);
+      console.log(`✔ Backed up existing ~/.openclaude.json to: ${globalConfigBackupPath}`);
+    }
+
+    try {
+      const globalConfig = JSON.parse(fs.readFileSync(globalConfigPath, "utf-8"));
+      globalConfig.providerProfiles = Array.isArray(globalConfig.providerProfiles) ? globalConfig.providerProfiles : [];
+
+      const existingIndex = globalConfig.providerProfiles.findIndex(p => p.id === PROVIDER_ID || p.name === PROVIDER_NAME);
+      const unlockedProfile = {
+        id: PROVIDER_ID,
+        name: PROVIDER_NAME,
+        provider: "openai",
+        baseUrl,
+        model: targetModel,
+        apiKey: "opencode-unlocked"
+      };
+
+      if (existingIndex !== -1) {
+        globalConfig.providerProfiles[existingIndex] = unlockedProfile;
+      } else {
+        globalConfig.providerProfiles.push(unlockedProfile);
+      }
+
+      globalConfig.activeProviderProfileId = PROVIDER_ID;
+
+      // Register model options cache for picker UI
+      globalConfig.openaiAdditionalModelOptionsCacheByProfile = globalConfig.openaiAdditionalModelOptionsCacheByProfile || {};
+      globalConfig.openaiAdditionalModelOptionsCacheByProfile[PROVIDER_ID] = [
+        { value: "muse-spark-1.3-contributor", label: "Muse Spark 1.3 Contributor", description: "Default Reasoning Model" },
+        { value: "space-bunny", label: "Space Bunny", description: "Lightweight Chat & Code Model" },
+        { value: "deepseek-v4-flash", label: "DeepSeek V4 Flash", description: "High-Throughput Model" },
+        { value: "glm-5.3", label: "GLM 5.3", description: "General Language Model" }
+      ];
+
+      fs.writeFileSync(globalConfigPath, JSON.stringify(globalConfig, null, 2), "utf-8");
+      console.log(`✔ Updated ~/.openclaude.json (active provider set to: ${PROVIDER_NAME})`);
+    } catch (err) {
+      console.warn(`⚠ Could not update ~/.openclaude.json: ${err.message}`);
+    }
+  }
+
+  // 2. Configure legacy mirror ~/.openclaude/.openclaude-profile.json
+  if (!fs.existsSync(openclaudeDir)) {
+    fs.mkdirSync(openclaudeDir, { recursive: true });
+  }
+
+  if (fs.existsSync(profilePath) && !fs.existsSync(profileBackupPath)) {
+    fs.copyFileSync(profilePath, profileBackupPath);
+    console.log(`✔ Backed up existing .openclaude-profile.json to: ${profileBackupPath}`);
+  }
+
+  const legacyProfile = {
+    profile: "openai",
+    env: {
+      OPENAI_BASE_URL: baseUrl,
+      OPENAI_MODEL: targetModel,
+      OPENAI_API_KEY: "opencode-unlocked"
+    },
+    displayName: PROVIDER_NAME,
+    createdAt: new Date().toISOString()
+  };
+
+  fs.writeFileSync(profilePath, JSON.stringify(legacyProfile, null, 2), "utf-8");
+  console.log(`✔ OpenClaude provider profile configured successfully!`);
+  console.log(`  Provider Name  : ${PROVIDER_NAME}`);
+  console.log(`  OPENAI_BASE_URL: ${baseUrl}`);
+  console.log(`  OPENAI_MODEL   : ${targetModel}`);
+
+  return legacyProfile;
+}
+
+export function restoreOriginalProfile() {
+  let restored = false;
+  if (fs.existsSync(globalConfigBackupPath)) {
+    fs.copyFileSync(globalConfigBackupPath, globalConfigPath);
+    console.log(`✔ Restored ~/.openclaude.json from: ${globalConfigBackupPath}`);
+    restored = true;
+  }
+  if (fs.existsSync(profileBackupPath)) {
+    fs.copyFileSync(profileBackupPath, profilePath);
+    console.log(`✔ Restored .openclaude-profile.json from: ${profileBackupPath}`);
+    restored = true;
+  }
+  return restored;
+}
+
+// CLI Execution
+if (process.argv[1] && process.argv[1].endsWith("setup.mjs")) {
+  const arg = process.argv[2];
+  if (arg === "restore" || arg === "--restore") {
+    restoreOriginalProfile();
+  } else {
+    configureOpenClaudeProfile({ model: arg });
+  }
+}
diff --git a/commands/status.md b/commands/status.md
new file mode 100644
index 0000000..d2110a6
--- /dev/null
+++ b/commands/status.md
@@ -0,0 +1,8 @@
+---
+description: Check Opencode Unlocked status, active model, and quota statistics
+argument-hint: ""
+---
+
+Check the status of the Opencode Unlocked bridge and show connection health:
+1. Query `http://127.0.0.1:8788/status` or execute `${CLAUDE_PLUGIN_ROOT}/bin/opencode-unlocked status`.
+2. Report the active default model (Muse Spark 1.3 Contributor), upstream connectivity, and session request/token usage statistics.
diff --git a/commands/unlocked.md b/commands/unlocked.md
new file mode 100644
index 0000000..bd8c0e8
--- /dev/null
+++ b/commands/unlocked.md
@@ -0,0 +1,11 @@
+---
+description: Manage Opencode Unlocked bridge, switch models, or configure keys
+argument-hint: "[models|switch <model>|key <key>|test]"
+---
+
+Manage Opencode Unlocked:
+- If argument is `models`: Run `${CLAUDE_PLUGIN_ROOT}/bin/opencode-unlocked models` to list all 36+ available models.
+- If argument is `switch <model>`: Run `${CLAUDE_PLUGIN_ROOT}/bin/opencode-unlocked switch $1` to change the default model.
+- If argument is `key <key>`: Run `${CLAUDE_PLUGIN_ROOT}/bin/opencode-unlocked set-key $1` to securely store your OpenCode API key.
+- If argument is `test`: Run `${CLAUDE_PLUGIN_ROOT}/bin/opencode-unlocked test` to test live connectivity to Muse Spark 1.3 Contributor and Space Bunny.
+- Otherwise, run `${CLAUDE_PLUGIN_ROOT}/bin/opencode-unlocked status`.
diff --git a/hooks/hooks.json b/hooks/hooks.json
new file mode 100644
index 0000000..b8950fb
--- /dev/null
+++ b/hooks/hooks.json
@@ -0,0 +1,16 @@
+{
+  "description": "Opencode Unlocked session lifecycle hooks",
+  "hooks": {
+    "SessionStart": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "bash \"${CLAUDE_PLUGIN_ROOT}/hooks/session-start.sh\"",
+            "async": false
+          }
+        ]
+      }
+    ]
+  }
+}
diff --git a/hooks/session-start.sh b/hooks/session-start.sh
new file mode 100755
index 0000000..86eb717
--- /dev/null
+++ b/hooks/session-start.sh
@@ -0,0 +1,56 @@
+#!/usr/bin/env bash
+set -e
+
+# Resolve plugin root if not set
+PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-$(cd "$(dirname "$0")/.." && pwd)}"
+PID_FILE="${PLUGIN_ROOT}/.gateway.pid"
+PORT="${OPENCODE_UNLOCKED_PORT:-8788}"
+GATEWAY_URL="http://127.0.0.1:${PORT}"
+
+# Check if gateway is already responding
+is_gateway_running() {
+  curl -s -f -m 1 "${GATEWAY_URL}/health" >/dev/null 2>&1
+}
+
+if ! is_gateway_running; then
+  # If a dead pid file exists, clean it up
+  if [ -f "$PID_FILE" ]; then
+    OLD_PID=$(cat "$PID_FILE" 2>/dev/null || true)
+    if [ -n "$OLD_PID" ] && ! kill -0 "$OLD_PID" 2>/dev/null; then
+      rm -f "$PID_FILE"
+    fi
+  fi
+
+  # Start daemon
+  nohup node "${PLUGIN_ROOT}/server/gateway.mjs" >/tmp/opencode-unlocked.log 2>&1 &
+  DAEMON_PID=$!
+  echo "$DAEMON_PID" > "$PID_FILE"
+
+  # Wait up to 3 seconds for gateway to become healthy
+  for _ in {1..30}; do
+    if is_gateway_running; then
+      break
+    fi
+    sleep 0.1
+  done
+fi
+
+# If OpenClaude passed an environment file, inject provider env vars
+if [ -n "$CLAUDE_ENV_FILE" ]; then
+  cat << EOF >> "$CLAUDE_ENV_FILE"
+export OPENAI_BASE_URL="${GATEWAY_URL}/v1"
+if [ -z "\$OPENAI_MODEL" ]; then
+  export OPENAI_MODEL="muse-spark-1.3-contributor"
+fi
+EOF
+fi
+
+# Emit JSON output for OpenClaude SessionStart hook
+cat << EOF
+{
+  "hookSpecificOutput": {
+    "additionalContext": "Opencode Unlocked active at ${GATEWAY_URL}/v1. Default model: muse-spark-1.3-contributor (Space Bunny and 36+ models available)."
+  }
+}
+EOF
+exit 0
diff --git a/package.json b/package.json
new file mode 100644
index 0000000..8d1a8b9
--- /dev/null
+++ b/package.json
@@ -0,0 +1,27 @@
+{
+  "name": "opencode-unlocked",
+  "version": "1.0.0",
+  "description": "Seamless OpenClaude provider plugin and authentic harness bridge for OpenCode Go models.",
+  "type": "module",
+  "main": "server/gateway.mjs",
+  "bin": {
+    "opencode-unlocked": "./bin/opencode-unlocked"
+  },
+  "scripts": {
+    "start": "node server/gateway.mjs",
+    "setup": "node bin/setup.mjs",
+    "test": "node test/session.test.mjs && node test/translator.test.mjs && node test/gateway.test.mjs",
+    "test:live": "node test/live-models.test.mjs"
+  },
+  "keywords": [
+    "openclaude",
+    "openclaude-plugin",
+    "opencode",
+    "opencode-go",
+    "muse-spark",
+    "space-bunny",
+    "ai-bridge"
+  ],
+  "author": "Opencode Unlocked Contributors",
+  "license": "MIT"
+}
diff --git a/server/catalog.mjs b/server/catalog.mjs
new file mode 100644
index 0000000..d84c5c8
--- /dev/null
+++ b/server/catalog.mjs
@@ -0,0 +1,115 @@
+import { CONFIG, getApiKey } from "./config.mjs";
+import { generateSessionId } from "./session.mjs";
+import crypto from "node:crypto";
+
+class ModelCatalog {
+  constructor() {
+    this.models = [];
+    this.lastFetched = 0;
+    this.ttlMs = 15 * 60 * 1000; // 15 minutes cache
+  }
+
+  /**
+   * Normalizes model ID by stripping any plugin or harness prefixes.
+   */
+  normalizeModelId(modelId) {
+    if (!modelId || modelId === "default" || modelId === "auto") {
+      return CONFIG.DEFAULT_MODEL;
+    }
+    return modelId
+      .replace(/^opencode-unlocked-/, "")
+      .replace(/^opencode-go-/, "")
+      .trim();
+  }
+
+  /**
+   * Fetches models dynamically from upstream OpenCode Go /models endpoint.
+   */
+  async getModels(forceRefresh = false) {
+    const now = Date.now();
+    if (!forceRefresh && this.models.length > 0 && now - this.lastFetched < this.ttlMs) {
+      return this.models;
+    }
+
+    const apiKey = getApiKey();
+    if (!apiKey) {
+      // Return static fallback if no key is configured yet
+      return this.getFallbackModels();
+    }
+
+    const sessionId = generateSessionId();
+    const requestId = crypto.randomUUID();
+
+    try {
+      const resp = await fetch(`${CONFIG.UPSTREAM_BASE_URL}/models`, {
+        method: "GET",
+        headers: {
+          "Authorization": `Bearer ${apiKey}`,
+          "User-Agent": CONFIG.USER_AGENT,
+          "x-opencode-client": CONFIG.CLIENT,
+          "x-opencode-project": CONFIG.PROJECT,
+          "x-opencode-session": sessionId,
+          "x-opencode-session-id": sessionId,
+          "x-session-affinity": sessionId,
+          "X-Session-Id": sessionId,
+          "x-opencode-request": requestId,
+          "Accept": "application/json"
+        }
+      });
+
+      if (!resp.ok) {
+        console.error(`[Opencode Unlocked] Failed to fetch upstream models: ${resp.status}`);
+        return this.models.length > 0 ? this.models : this.getFallbackModels();
+      }
+
+      const data = await resp.json();
+      const rawList = Array.isArray(data) ? data : (Array.isArray(data.data) ? data.data : []);
+
+      this.models = rawList.map((m) => {
+        const id = typeof m === "string" ? m : m.id;
+        const name = typeof m === "string" ? m : (m.name || m.id);
+        return {
+          id,
+          object: "model",
+          created: 1700000000,
+          owned_by: "opencode",
+          name,
+          is_default: id === CONFIG.DEFAULT_MODEL
+        };
+      });
+
+      // Ensure default model is in the catalog
+      if (!this.models.some((m) => m.id === CONFIG.DEFAULT_MODEL)) {
+        this.models.unshift({
+          id: CONFIG.DEFAULT_MODEL,
+          object: "model",
+          created: 1700000000,
+          owned_by: "opencode",
+          name: "Muse Spark 1.3 Contributor",
+          is_default: true
+        });
+      }
+
+      this.lastFetched = now;
+      return this.models;
+    } catch (err) {
+      console.error(`[Opencode Unlocked] Error fetching models: ${err.message}`);
+      return this.models.length > 0 ? this.models : this.getFallbackModels();
+    }
+  }
+
+  getFallbackModels() {
+    return [
+      { id: "muse-spark-1.3-contributor", name: "Muse Spark 1.3 Contributor", is_default: true },
+      { id: "space-bunny", name: "Space Bunny", is_default: false },
+      { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash", is_default: false },
+      { id: "glm-5.3", name: "GLM 5.3", is_default: false },
+      { id: "gpt-5.6-luna", name: "GPT 5.6 Luna", is_default: false },
+      { id: "kimi-k2.5", name: "Kimi K2.5", is_default: false },
+      { id: "minimax-01", name: "MiniMax 01", is_default: false },
+      { id: "qwen3.5-max", name: "Qwen 3.5 Max", is_default: false }
+    ].map(m => ({ ...m, object: "model", created: 1700000000, owned_by: "opencode" }));
+  }
+}
+
+export const catalog = new ModelCatalog();
diff --git a/server/config.mjs b/server/config.mjs
new file mode 100644
index 0000000..30fb5a6
--- /dev/null
+++ b/server/config.mjs
@@ -0,0 +1,100 @@
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+function parseEnvFile(filePath) {
+  if (!fs.existsSync(filePath)) return {};
+  try {
+    const content = fs.readFileSync(filePath, "utf-8");
+    const result = {};
+    for (const line of content.split("\n")) {
+      const trimmed = line.trim();
+      if (!trimmed || trimmed.startsWith("#")) continue;
+      const eqIdx = trimmed.indexOf("=");
+      if (eqIdx !== -1) {
+        const key = trimmed.slice(0, eqIdx).trim();
+        let val = trimmed.slice(eqIdx + 1).trim();
+        if ((val.startsWith('"') && val.endsWith('"')) || (val.startsWith("'") && val.endsWith("'"))) {
+          val = val.slice(1, -1);
+        }
+        result[key] = val;
+      }
+    }
+    return result;
+  } catch {
+    return {};
+  }
+}
+
+export function getUserConfigPath() {
+  const home = os.homedir();
+  const primaryDir = path.join(home, ".openclaude");
+  if (!fs.existsSync(primaryDir)) {
+    try {
+      fs.mkdirSync(primaryDir, { recursive: true });
+    } catch {
+      // ignore
+    }
+  }
+  return path.join(primaryDir, "opencode-unlocked.json");
+}
+
+export function loadUserConfig() {
+  const configPath = getUserConfigPath();
+  if (fs.existsSync(configPath)) {
+    try {
+      return JSON.parse(fs.readFileSync(configPath, "utf-8"));
+    } catch {
+      return {};
+    }
+  }
+  return {};
+}
+
+export function saveUserConfig(data) {
+  const configPath = getUserConfigPath();
+  const dir = path.dirname(configPath);
+  if (!fs.existsSync(dir)) {
+    fs.mkdirSync(dir, { recursive: true });
+  }
+  const current = loadUserConfig();
+  const updated = { ...current, ...data };
+  fs.writeFileSync(configPath, JSON.stringify(updated, null, 2), "utf-8");
+  return updated;
+}
+
+export function getApiKey() {
+  // 1. Process environment variables
+  if (process.env.OPENCODE_UNLOCKED_API_KEY) return process.env.OPENCODE_UNLOCKED_API_KEY;
+  if (process.env.OPENCODE_API_KEY) return process.env.OPENCODE_API_KEY;
+  if (process.env.GO_KEY) return process.env.GO_KEY;
+
+  // 2. User config (~/.openclaude/opencode-unlocked.json)
+  const userConfig = loadUserConfig();
+  if (userConfig.apiKey) return userConfig.apiKey;
+  if (userConfig.GO_KEY) return userConfig.GO_KEY;
+
+  // 3. Optional local .env (checked in CWD and repo root, strictly ignored by git)
+  const cwdEnv = parseEnvFile(path.join(process.cwd(), ".env"));
+  if (cwdEnv.OPENCODE_UNLOCKED_API_KEY) return cwdEnv.OPENCODE_UNLOCKED_API_KEY;
+  if (cwdEnv.OPENCODE_API_KEY) return cwdEnv.OPENCODE_API_KEY;
+  if (cwdEnv.GO_KEY) return cwdEnv.GO_KEY;
+
+  const repoEnv = parseEnvFile(path.join(import.meta.dirname, "..", ".env"));
+  if (repoEnv.OPENCODE_UNLOCKED_API_KEY) return repoEnv.OPENCODE_UNLOCKED_API_KEY;
+  if (repoEnv.OPENCODE_API_KEY) return repoEnv.OPENCODE_API_KEY;
+  if (repoEnv.GO_KEY) return repoEnv.GO_KEY;
+
+  return null;
+}
+
+export const CONFIG = {
+  PORT: parseInt(process.env.OPENCODE_UNLOCKED_PORT || "8788", 10),
+  HOST: process.env.OPENCODE_UNLOCKED_HOST || "127.0.0.1",
+  UPSTREAM_BASE_URL: (process.env.OPENCODE_BASE_URL || "https://opencode.ai/zen/go/v1").replace(/\/+$/, ""),
+  DEFAULT_MODEL: process.env.OPENCODE_UNLOCKED_DEFAULT_MODEL || "muse-spark-1.3-contributor",
+  USER_AGENT: "opencode/stable/2.0.24/cli",
+  CLIENT: "cli",
+  PROJECT: "default",
+  VERSION: "1.0.0"
+};
diff --git a/server/gateway.mjs b/server/gateway.mjs
new file mode 100644
index 0000000..e0e6048
--- /dev/null
+++ b/server/gateway.mjs
@@ -0,0 +1,395 @@
+import http from "node:http";
+import crypto from "node:crypto";
+import { CONFIG, getApiKey } from "./config.mjs";
+import { sessionAffinity, generateSessionId } from "./session.mjs";
+import { catalog } from "./catalog.mjs";
+import { usageTracker } from "./usage.mjs";
+import {
+  requiresResponsesApi,
+  chatCompletionsToResponsesBody,
+  responsesToChatCompletionsResponse,
+  translateResponsesEventToOpenAiChunk
+} from "./translator.mjs";
+
+function sendJson(res, statusCode, data) {
+  const jsonStr = JSON.stringify(data);
+  res.writeHead(statusCode, {
+    "Content-Type": "application/json; charset=utf-8",
+    "Content-Length": Buffer.byteLength(jsonStr),
+    "Access-Control-Allow-Origin": "*",
+    "Access-Control-Allow-Methods": "GET, POST, OPTIONS, HEAD",
+    "Access-Control-Allow-Headers": "*"
+  });
+  res.end(jsonStr);
+}
+
+function parseBody(req) {
+  return new Promise((resolve, reject) => {
+    let body = "";
+    req.on("data", (chunk) => {
+      body += chunk;
+      if (body.length > 50 * 1024 * 1024) { // 50MB protection
+        reject(new Error("Request body too large"));
+      }
+    });
+    req.on("end", () => {
+      if (!body) return resolve({});
+      try {
+        resolve(JSON.parse(body));
+      } catch (err) {
+        reject(new Error(`Invalid JSON body: ${err.message}`));
+      }
+    });
+    req.on("error", reject);
+  });
+}
+
+export function createGatewayServer() {
+  const server = http.createServer(async (req, res) => {
+    // Enable CORS preflight
+    if (req.method === "OPTIONS") {
+      res.writeHead(204, {
+        "Access-Control-Allow-Origin": "*",
+        "Access-Control-Allow-Methods": "GET, POST, OPTIONS, HEAD",
+        "Access-Control-Allow-Headers": "*",
+        "Access-Control-Max-Age": "86400"
+      });
+      res.end();
+      return;
+    }
+
+    const parsedUrl = new URL(req.url, `http://${req.headers.host || "127.0.0.1"}`);
+    const pathname = parsedUrl.pathname.replace(/\/+$/, "") || "/";
+
+    // Health Check
+    if (pathname === "/health" || pathname === "/") {
+      const apiKey = getApiKey();
+      sendJson(res, 200, {
+        status: "ok",
+        name: "Opencode Unlocked Gateway",
+        version: CONFIG.VERSION,
+        hasKey: !!apiKey,
+        defaultModel: CONFIG.DEFAULT_MODEL
+      });
+      return;
+    }
+
+    // Status / Quota
+    if (pathname === "/status") {
+      const apiKey = getApiKey();
+      const upstreamUsage = await usageTracker.getUpstreamUsage();
+      sendJson(res, 200, {
+        status: "ok",
+        keyConfigured: !!apiKey,
+        defaultModel: CONFIG.DEFAULT_MODEL,
+        stats: usageTracker.getStatus(),
+        upstreamUsage
+      });
+      return;
+    }
+
+    // Models List (/v1/models and /models)
+    if (req.method === "GET" && (pathname === "/v1/models" || pathname === "/models")) {
+      try {
+        const models = await catalog.getModels();
+        sendJson(res, 200, {
+          object: "list",
+          data: models
+        });
+      } catch (err) {
+        sendJson(res, 500, { error: { message: err.message, type: "internal_error" } });
+      }
+      return;
+    }
+
+    // Chat Completions (/v1/chat/completions and /chat/completions)
+    if (req.method === "POST" && (pathname === "/v1/chat/completions" || pathname === "/chat/completions")) {
+      await handleChatCompletions(req, res);
+      return;
+    }
+
+    // Direct Responses endpoint (/v1/responses and /responses)
+    if (req.method === "POST" && (pathname === "/v1/responses" || pathname === "/responses")) {
+      await handleDirectResponses(req, res);
+      return;
+    }
+
+    sendJson(res, 404, {
+      error: {
+        message: `Endpoint not found: ${req.method} ${pathname}`,
+        type: "invalid_request_error"
+      }
+    });
+  });
+
+  return server;
+}
+
+async function handleChatCompletions(req, res) {
+  let body;
+  try {
+    body = await parseBody(req);
+  } catch (err) {
+    sendJson(res, 400, { error: { message: err.message, type: "invalid_request_error" } });
+    return;
+  }
+
+  // Resolve API Key
+  let apiKey = getApiKey();
+  const authHeader = req.headers.authorization;
+  if (authHeader && authHeader.startsWith("Bearer ")) {
+    const passedKey = authHeader.slice(7).trim();
+    if (passedKey.startsWith("oc_sk_")) {
+      apiKey = passedKey;
+    }
+  }
+
+  if (!apiKey) {
+    sendJson(res, 401, {
+      error: {
+        message: "No OpenCode API key configured. Provide OPENCODE_UNLOCKED_API_KEY, GO_KEY, or run `opencode-unlocked key <key>`.",
+        type: "authentication_error"
+      }
+    });
+    return;
+  }
+
+  const rawModel = body.model || CONFIG.DEFAULT_MODEL;
+  const normalizedModel = catalog.normalizeModelId(rawModel);
+  body.model = normalizedModel;
+
+  // Derive Session ID with Affinity
+  const clientIdentifier =
+    req.headers["x-session-id"] ||
+    req.headers["x-opencode-session"] ||
+    req.headers["x-conversation-id"] ||
+    req.socket.remoteAddress ||
+    "default";
+  const sessionId = sessionAffinity.getOrCreate(clientIdentifier);
+  const requestId = crypto.randomUUID();
+
+  const useResponses = requiresResponsesApi(normalizedModel);
+  const isStream = !!body.stream;
+
+  const harnessHeaders = {
+    "Authorization": `Bearer ${apiKey}`,
+    "User-Agent": CONFIG.USER_AGENT,
+    "x-opencode-client": CONFIG.CLIENT,
+    "x-opencode-project": CONFIG.PROJECT,
+    "x-opencode-session": sessionId,
+    "x-opencode-session-id": sessionId,
+    "x-session-affinity": sessionId,
+    "X-Session-Id": sessionId,
+    "x-opencode-request": requestId,
+    "Content-Type": "application/json",
+    "Accept": isStream ? "text/event-stream" : "application/json"
+  };
+
+  try {
+    if (useResponses) {
+      // Protocol Translation: ChatCompletions -> Responses API
+      const responsesPayload = chatCompletionsToResponsesBody(body);
+      const upstreamResp = await fetch(`${CONFIG.UPSTREAM_BASE_URL}/responses`, {
+        method: "POST",
+        headers: harnessHeaders,
+        body: JSON.stringify(responsesPayload)
+      });
+
+      if (!upstreamResp.ok) {
+        const errorText = await upstreamResp.text();
+        let parsedErr;
+        try {
+          parsedErr = JSON.parse(errorText);
+        } catch {
+          parsedErr = { error: { message: errorText, type: "upstream_error" } };
+        }
+        sendJson(res, upstreamResp.status, parsedErr);
+        return;
+      }
+
+      if (isStream) {
+        res.writeHead(200, {
+          "Content-Type": "text/event-stream; charset=utf-8",
+          "Cache-Control": "no-cache",
+          "Connection": "keep-alive",
+          "Access-Control-Allow-Origin": "*"
+        });
+
+        const state = { id: "", model: normalizedModel };
+        const reader = upstreamResp.body.getReader();
+        const decoder = new TextDecoder("utf-8");
+        let buffer = "";
+
+        while (true) {
+          const { done, value } = await reader.read();
+          if (done) break;
+
+          buffer += decoder.decode(value, { stream: true });
+          const parts = buffer.split("\n\n");
+          buffer = parts.pop() || "";
+
+          for (const block of parts) {
+            const trimmed = block.trim();
+            if (!trimmed) continue;
+            const convertedChunk = translateResponsesEventToOpenAiChunk(trimmed, state);
+            if (convertedChunk) {
+              res.write(convertedChunk);
+            }
+          }
+        }
+
+        if (buffer.trim()) {
+          const convertedChunk = translateResponsesEventToOpenAiChunk(buffer.trim(), state);
+          if (convertedChunk) res.write(convertedChunk);
+        }
+
+        res.write("data: [DONE]\n\n");
+        res.end();
+        usageTracker.recordUsage(state.usage || { prompt_tokens: 0, completion_tokens: 0, total_tokens: 0 });
+      } else {
+        const responseData = await upstreamResp.json();
+        const chatCompletionData = responsesToChatCompletionsResponse(responseData, normalizedModel);
+        usageTracker.recordUsage(chatCompletionData.usage);
+        sendJson(res, 200, chatCompletionData);
+      }
+    } else {
+      // Standard Chat Completions Pass-Through
+      const upstreamResp = await fetch(`${CONFIG.UPSTREAM_BASE_URL}/chat/completions`, {
+        method: "POST",
+        headers: harnessHeaders,
+        body: JSON.stringify(body)
+      });
+
+      if (!upstreamResp.ok) {
+        const errorText = await upstreamResp.text();
+        let parsedErr;
+        try {
+          parsedErr = JSON.parse(errorText);
+        } catch {
+          parsedErr = { error: { message: errorText, type: "upstream_error" } };
+        }
+        sendJson(res, upstreamResp.status, parsedErr);
+        return;
+      }
+
+      if (isStream) {
+        res.writeHead(200, {
+          "Content-Type": "text/event-stream; charset=utf-8",
+          "Cache-Control": "no-cache",
+          "Connection": "keep-alive",
+          "Access-Control-Allow-Origin": "*"
+        });
+
+        const reader = upstreamResp.body.getReader();
+        while (true) {
+          const { done, value } = await reader.read();
+          if (done) break;
+          res.write(value);
+        }
+        res.end();
+        usageTracker.recordUsage({ prompt_tokens: 0, completion_tokens: 0, total_tokens: 0 });
+      } else {
+        const json = await upstreamResp.json();
+        usageTracker.recordUsage(json.usage);
+        sendJson(res, 200, json);
+      }
+    }
+  } catch (err) {
+    console.error("[Opencode Unlocked] Upstream relay error:", err);
+    sendJson(res, 502, {
+      error: {
+        message: `Gateway relay error: ${err.message}`,
+        type: "bad_gateway"
+      }
+    });
+  }
+}
+
+async function handleDirectResponses(req, res) {
+  let body;
+  try {
+    body = await parseBody(req);
+  } catch (err) {
+    sendJson(res, 400, { error: { message: err.message, type: "invalid_request_error" } });
+    return;
+  }
+
+  const apiKey = getApiKey();
+  if (!apiKey) {
+    sendJson(res, 401, { error: { message: "No OpenCode API key configured.", type: "authentication_error" } });
+    return;
+  }
+
+  const rawModel = body.model || CONFIG.DEFAULT_MODEL;
+  body.model = catalog.normalizeModelId(rawModel);
+
+  const sessionId = generateSessionId();
+  const isStream = !!body.stream;
+
+  const harnessHeaders = {
+    "Authorization": `Bearer ${apiKey}`,
+    "User-Agent": CONFIG.USER_AGENT,
+    "x-opencode-client": CONFIG.CLIENT,
+    "x-opencode-project": CONFIG.PROJECT,
+    "x-opencode-session": sessionId,
+    "x-opencode-session-id": sessionId,
+    "x-session-affinity": sessionId,
+    "X-Session-Id": sessionId,
+    "x-opencode-request": crypto.randomUUID(),
+    "Content-Type": "application/json",
+    "Accept": isStream ? "text/event-stream" : "application/json"
+  };
+
+  try {
+    const upstreamResp = await fetch(`${CONFIG.UPSTREAM_BASE_URL}/responses`, {
+      method: "POST",
+      headers: harnessHeaders,
+      body: JSON.stringify(body)
+    });
+
+    if (!upstreamResp.ok) {
+      const errText = await upstreamResp.text();
+      let json;
+      try { json = JSON.parse(errText); } catch { json = { error: { message: errText } }; }
+      sendJson(res, upstreamResp.status, json);
+      return;
+    }
+
+    if (isStream) {
+      res.writeHead(200, {
+        "Content-Type": "text/event-stream; charset=utf-8",
+        "Cache-Control": "no-cache",
+        "Connection": "keep-alive"
+      });
+      const reader = upstreamResp.body.getReader();
+      while (true) {
+        const { done, value } = await reader.read();
+        if (done) break;
+        res.write(value);
+      }
+      res.end();
+    } else {
+      const json = await upstreamResp.json();
+      sendJson(res, 200, json);
+    }
+  } catch (err) {
+    sendJson(res, 502, { error: { message: err.message, type: "bad_gateway" } });
+  }
+}
+
+// CLI Execution entrypoint
+if (process.argv[1] && process.argv[1].endsWith("gateway.mjs")) {
+  const server = createGatewayServer();
+  server.listen(CONFIG.PORT, CONFIG.HOST, () => {
+    console.log(`[Opencode Unlocked] Bridge Gateway running on http://${CONFIG.HOST}:${CONFIG.PORT}`);
+    console.log(`[Opencode Unlocked] Default model: ${CONFIG.DEFAULT_MODEL}`);
+    console.log(`[Opencode Unlocked] API key resolved: ${!!getApiKey()}`);
+  });
+
+  const shutdown = () => {
+    console.log("\n[Opencode Unlocked] Shutting down gateway...");
+    server.close(() => process.exit(0));
+  };
+  process.on("SIGINT", shutdown);
+  process.on("SIGTERM", shutdown);
+}
diff --git a/server/session.mjs b/server/session.mjs
new file mode 100644
index 0000000..41e9c88
--- /dev/null
+++ b/server/session.mjs
@@ -0,0 +1,89 @@
+import crypto from "node:crypto";
+
+const ID_LENGTH = 26;
+const BASE62_CHARS = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
+
+let lastTimestamp = 0;
+let counter = 0;
+
+/**
+ * OpenCode v2 descending identifier generator.
+ * Produces a time-ordered sortable 26-character identifier.
+ * First 12 characters: hex-encoded bitwise-inverted timestamp (48 bits).
+ * Remaining 14 characters: cryptographically random base62 characters.
+ */
+export function generateDescendingIdentifier(timestamp = Date.now()) {
+  if (timestamp !== lastTimestamp) {
+    lastTimestamp = timestamp;
+    counter = 0;
+  }
+  counter++;
+
+  const current = BigInt(timestamp) * 0x1000n + BigInt(counter);
+  const value = ~current; // Descending sort order
+  const timeHex = Array.from({ length: 6 }, (_, index) =>
+    Number((value >> BigInt(40 - 8 * index)) & 0xffn)
+      .toString(16)
+      .padStart(2, "0"),
+  ).join("");
+
+  const randomBytes = crypto.getRandomValues(new Uint8Array(ID_LENGTH - 12));
+  const randomChars = Array.from(randomBytes, (b) => BASE62_CHARS[b % 62]).join("");
+
+  return timeHex + randomChars;
+}
+
+/**
+ * Generates an authentic OpenCode v2 Session ID.
+ * Format: "ses_" + 26 chars = exactly 30 characters.
+ * Matches packages/schema/src/session-id.ts exactly.
+ */
+export function generateSessionId() {
+  return `ses_${generateDescendingIdentifier()}`;
+}
+
+/**
+ * Validates whether a given string matches the authentic OpenCode v2 Session ID format.
+ */
+export function isValidSessionId(id) {
+  if (typeof id !== "string" || id.length !== 30) return false;
+  return /^ses_[0-9a-f]{12}[0-9A-Za-z]{14}$/.test(id);
+}
+
+/**
+ * Session Affinity Cache
+ * Maps client conversation IDs (or IP/caller hashes) to persistent OpenCode session IDs
+ * to ensure prompt-cache affinity across turns.
+ */
+class SessionAffinityManager {
+  constructor(ttlMs = 4 * 60 * 60 * 1000) { // 4 hours default TTL
+    this.sessions = new Map();
+    this.ttlMs = ttlMs;
+  }
+
+  getOrCreate(clientIdentifier) {
+    const key = clientIdentifier || "default";
+    const now = Date.now();
+    const existing = this.sessions.get(key);
+
+    if (existing && now - existing.timestamp < this.ttlMs) {
+      existing.timestamp = now; // Slide TTL window
+      return existing.sessionId;
+    }
+
+    const sessionId = generateSessionId();
+    this.sessions.set(key, { sessionId, timestamp: now });
+    return sessionId;
+  }
+
+  set(clientIdentifier, sessionId) {
+    const key = clientIdentifier || "default";
+    this.sessions.set(key, { sessionId, timestamp: Date.now() });
+  }
+
+  clear() {
+    this.sessions.clear();
+  }
+}
+
+export const sessionAffinity = new SessionAffinityManager();
diff --git a/server/translator.mjs b/server/translator.mjs
new file mode 100644
index 0000000..a0b5298
--- /dev/null
+++ b/server/translator.mjs
@@ -0,0 +1,373 @@
+/**
+ * Protocol Translation Layer
+ * 
+ * OpenCode Go hosts models across two primary OpenAI API protocols:
+ * 1. Standard Chat Completions (/chat/completions): space-bunny, deepseek-v4-flash, glm-5.3, etc.
+ * 2. Responses API (/responses): muse-spark-1.3-contributor, gpt-5.6-luna, gpt-5.*, grok-*
+ * 
+ * OpenClaude sends standard /v1/chat/completions.
+ * This module transparently adapts requests and streaming responses for models requiring /responses.
+ */
+
+// Models that strictly mandate the OpenAI Responses API protocol (/responses)
+const RESPONSES_API_MODELS = new Set([
+  "muse-spark-1.3-contributor",
+  "gpt-5.6-luna",
+  "gpt-5.6-sol",
+  "gpt-5.6-terra",
+  "gpt-5.5",
+  "gpt-5.5-pro",
+  "gpt-5.4",
+  "gpt-5.4-pro",
+  "gpt-5.4-mini",
+  "gpt-5.4-nano",
+  "gpt-5.3-codex",
+  "gpt-5.3-codex-spark",
+  "gpt-5.2",
+  "gpt-5.2-codex",
+  "gpt-5.1",
+  "gpt-5.1-codex",
+  "gpt-5.1-codex-max",
+  "gpt-5.1-codex-mini",
+  "gpt-5",
+  "gpt-5-codex",
+  "gpt-5-nano",
+  "grok-4.5",
+  "grok-build-0.1",
+]);
+
+/**
+ * Checks whether a given model requires the /responses endpoint.
+ */
+export function requiresResponsesApi(modelName) {
+  if (!modelName) return false;
+  const normalized = modelName.toLowerCase().replace(/^(opencode-unlocked-|opencode-go-)/, "");
+  if (RESPONSES_API_MODELS.has(normalized)) return true;
+  // Prefix matching for gpt-5.* family if newer variants appear
+  if (normalized.startsWith("gpt-5") || normalized.startsWith("muse-spark")) return true;
+  return false;
+}
+
+/**
+ * Translates an OpenAI Chat Completions request body into an OpenAI Responses API body.
+ */
+export function chatCompletionsToResponsesBody(body) {
+  const { messages, input, model, stream, temperature, max_tokens, max_output_tokens, tools, tool_choice, stop } = body;
+
+  const rawMessages = Array.isArray(input) ? input : Array.isArray(messages) ? messages : [];
+
+  // Map messages to Responses API input format
+  const formattedInput = rawMessages.map((msg) => {
+    if (!msg || typeof msg !== "object") return msg;
+
+    // Direct Responses API items pass through
+    if (!msg.role && msg.type) return msg;
+
+    // Standard Chat Completion messages
+    const role = msg.role;
+    const content = msg.content;
+
+    if (role === "system" || role === "developer") {
+      return {
+        role: "system",
+        content: typeof content === "string" ? content : (Array.isArray(content) ? content.map(c => c.text || "").join("") : "")
+      };
+    }
+
+    if (role === "user") {
+      if (typeof content === "string") {
+        return { role: "user", content };
+      }
+      if (Array.isArray(content)) {
+        return {
+          role: "user",
+          content: content.map((part) => {
+            if (part.type === "text" || part.type === "input_text") {
+              return { type: "text", text: part.text || "" };
+            }
+            if (part.type === "image_url") {
+              return { type: "image_url", image_url: part.image_url };
+            }
+            return part;
+          })
+        };
+      }
+      return { role: "user", content: String(content || "") };
+    }
+
+    if (role === "assistant") {
+      const out = { role: "assistant" };
+      if (typeof content === "string") out.content = content;
+      if (Array.isArray(msg.tool_calls)) out.tool_calls = msg.tool_calls;
+      return out;
+    }
+
+    if (role === "tool") {
+      return {
+        role: "tool",
+        tool_call_id: msg.tool_call_id,
+        content: typeof content === "string" ? content : JSON.stringify(content || "")
+      };
+    }
+
+    return msg;
+  });
+
+  const responsesPayload = {
+    model: model.replace(/^(opencode-unlocked-|opencode-go-)/, ""),
+    input: formattedInput,
+    stream: !!stream,
+  };
+
+  if (typeof temperature === "number") responsesPayload.temperature = temperature;
+  if (typeof max_output_tokens === "number") responsesPayload.max_output_tokens = max_output_tokens;
+  else if (typeof max_tokens === "number") responsesPayload.max_output_tokens = max_tokens;
+
+  if (Array.isArray(tools)) {
+    responsesPayload.tools = tools.map((t) => {
+      if (t.type === "function" && t.function) {
+        return {
+          type: "function",
+          name: t.function.name,
+          description: t.function.description,
+          parameters: t.function.parameters,
+          ...(t.function.strict !== undefined ? { strict: t.function.strict } : {})
+        };
+      }
+      return t;
+    });
+  }
+
+  if (tool_choice) {
+    if (tool_choice === "auto" || tool_choice === "required" || tool_choice === "none") {
+      responsesPayload.tool_choice = tool_choice;
+    } else if (typeof tool_choice === "object" && tool_choice.type === "function" && tool_choice.function?.name) {
+      responsesPayload.tool_choice = { type: "function", name: tool_choice.function.name };
+    } else {
+      responsesPayload.tool_choice = tool_choice;
+    }
+  }
+
+  if (stop) responsesPayload.stop = stop;
+
+  return responsesPayload;
+}
+
+/**
+ * Translates a non-streaming Responses API JSON response to an OpenAI Chat Completions response.
+ */
+export function responsesToChatCompletionsResponse(responsesData, requestedModel) {
+  if (!responsesData || typeof responsesData !== "object") return responsesData;
+
+  const id = responsesData.id?.replace(/^resp_/, "chatcmpl_") || `chatcmpl_${Math.random().toString(36).slice(2)}`;
+  const model = responsesData.model || requestedModel;
+
+  let textContent = "";
+  const toolCalls = [];
+
+  const output = Array.isArray(responsesData.output) ? responsesData.output : [];
+  for (const item of output) {
+    if (item.type === "message") {
+      if (Array.isArray(item.content)) {
+        for (const part of item.content) {
+          if ((part.type === "output_text" || part.type === "text") && typeof part.text === "string") {
+            textContent += part.text;
+          }
+        }
+      } else if (typeof item.content === "string") {
+        textContent += item.content;
+      }
+    } else if (item.type === "function_call") {
+      toolCalls.push({
+        id: item.call_id || item.id,
+        type: "function",
+        function: {
+          name: item.name,
+          arguments: typeof item.arguments === "string" ? item.arguments : JSON.stringify(item.arguments || {})
+        }
+      });
+    }
+  }
+
+  const stopReason = responsesData.stop_reason;
+  let finishReason = "stop";
+  if (stopReason === "tool_call" || stopReason === "tool_calls" || toolCalls.length > 0) finishReason = "tool_calls";
+  else if (stopReason === "max_output_tokens" || stopReason === "length") finishReason = "length";
+  else if (stopReason === "content_filter") finishReason = "content_filter";
+
+  const message = {
+    role: "assistant",
+    content: textContent || (toolCalls.length ? null : "")
+  };
+  if (toolCalls.length > 0) {
+    message.tool_calls = toolCalls;
+  }
+
+  const usage = responsesData.usage ? {
+    prompt_tokens: responsesData.usage.input_tokens || 0,
+    completion_tokens: responsesData.usage.output_tokens || 0,
+    total_tokens: (responsesData.usage.input_tokens || 0) + (responsesData.usage.output_tokens || 0),
+    ...(responsesData.usage.input_tokens_details?.cached_tokens ? {
+      prompt_tokens_details: { cached_tokens: responsesData.usage.input_tokens_details.cached_tokens }
+    } : {})
+  } : undefined;
+
+  return {
+    id,
+    object: "chat.completion",
+    created: Math.floor(Date.now() / 1000),
+    model,
+    choices: [
+      {
+        index: 0,
+        message,
+        finish_reason: finishReason
+      }
+    ],
+    ...(usage ? { usage } : {})
+  };
+}
+
+/**
+ * Transforms an SSE event line or block from Responses API format
+ * into OpenAI chat.completion.chunk SSE format.
+ */
+export function translateResponsesEventToOpenAiChunk(eventString, state = { id: "", model: "" }) {
+  if (!eventString || typeof eventString !== "string") return null;
+
+  const lines = eventString.split("\n");
+  let eventType = "";
+  let dataStr = "";
+
+  for (const line of lines) {
+    if (line.startsWith("event:")) {
+      eventType = line.slice(6).trim();
+    } else if (line.startsWith("data:")) {
+      dataStr = line.slice(5).trim();
+    }
+  }
+
+  if (!dataStr) return null;
+  if (dataStr === "[DONE]") return "data: [DONE]\n\n";
+
+  let data;
+  try {
+    data = JSON.parse(dataStr);
+  } catch {
+    return null;
+  }
+
+  const respObj = data.response || {};
+  const id = respObj.id || data.id || state.id || `chatcmpl_${Math.random().toString(36).slice(2)}`;
+  const model = respObj.model || data.model || state.model || "opencode-unlocked";
+
+  state.id = id;
+  state.model = model;
+
+  const baseChunk = {
+    id,
+    object: "chat.completion.chunk",
+    created: Math.floor(Date.now() / 1000),
+    model,
+    choices: []
+  };
+
+  // Text Delta
+  if (
+    eventType === "response.output_text.delta" ||
+    eventType === "response.text.delta" ||
+    data.type === "response.output_text.delta"
+  ) {
+    const text = data.delta ?? data.text ?? data.output_text_delta ?? "";
+    if (text) {
+      baseChunk.choices.push({
+        index: 0,
+        delta: { content: text },
+        finish_reason: null
+      });
+      return `data: ${JSON.stringify(baseChunk)}\n\n`;
+    }
+  }
+
+  // Function Call Added
+  if (
+    (eventType === "response.output_item.added" || data.type === "response.output_item.added") &&
+    data.item?.type === "function_call"
+  ) {
+    baseChunk.choices.push({
+      index: 0,
+      delta: {
+        tool_calls: [
+          {
+            index: 0,
+            id: data.item.id || data.item.call_id,
+            type: "function",
+            function: {
+              name: data.item.name || "",
+              arguments: ""
+            }
+          }
+        ]
+      },
+      finish_reason: null
+    });
+    return `data: ${JSON.stringify(baseChunk)}\n\n`;
+  }
+
+  // Function Call Arguments Delta
+  if (
+    eventType === "response.function_call_arguments.delta" ||
+    data.type === "response.function_call_arguments.delta"
+  ) {
+    const argsDelta = data.delta ?? data.arguments_delta ?? "";
+    if (argsDelta) {
+      baseChunk.choices.push({
+        index: 0,
+        delta: {
+          tool_calls: [
+            {
+              index: 0,
+              function: {
+                arguments: argsDelta
+              }
+            }
+          ]
+        },
+        finish_reason: null
+      });
+      return `data: ${JSON.stringify(baseChunk)}\n\n`;
+    }
+  }
+
+  // Response Completed
+  if (eventType === "response.completed" || data.type === "response.completed") {
+    const sr = respObj.stop_reason ?? data.stop_reason;
+    let fr = "stop";
+    if (sr === "tool_call" || sr === "tool_calls") fr = "tool_calls";
+    else if (sr === "max_output_tokens" || sr === "length") fr = "length";
+    else if (sr === "content_filter") fr = "content_filter";
+
+    baseChunk.choices.push({
+      index: 0,
+      delta: {},
+      finish_reason: fr
+    });
+
+    const u = respObj.usage ?? data.response?.usage ?? data.usage;
+    if (u) {
+      baseChunk.usage = {
+        prompt_tokens: u.input_tokens || 0,
+        completion_tokens: u.output_tokens || 0,
+        total_tokens: (u.input_tokens || 0) + (u.output_tokens || 0),
+        ...(u.input_tokens_details?.cached_tokens ? {
+          prompt_tokens_details: { cached_tokens: u.input_tokens_details.cached_tokens }
+        } : {})
+      };
+      state.usage = baseChunk.usage;
+    }
+
+    return `data: ${JSON.stringify(baseChunk)}\n\n`;
+  }
+
+  return null;
+}
diff --git a/server/usage.mjs b/server/usage.mjs
new file mode 100644
index 0000000..f6b0753
--- /dev/null
+++ b/server/usage.mjs
@@ -0,0 +1,64 @@
+import { CONFIG, getApiKey } from "./config.mjs";
+import { generateSessionId } from "./session.mjs";
+import crypto from "node:crypto";
+
+class UsageTracker {
+  constructor() {
+    this.sessionRequests = 0;
+    this.sessionTokens = { input: 0, output: 0, total: 0 };
+    this.lastChecked = 0;
+    this.cachedUpstreamUsage = null;
+  }
+
+  recordUsage(usage) {
+    if (!usage) return;
+    this.sessionRequests++;
+    this.sessionTokens.input += usage.prompt_tokens || usage.input_tokens || 0;
+    this.sessionTokens.output += usage.completion_tokens || usage.output_tokens || 0;
+    this.sessionTokens.total += usage.total_tokens || 0;
+  }
+
+  async getUpstreamUsage() {
+    const apiKey = getApiKey();
+    if (!apiKey) return null;
+
+    const now = Date.now();
+    if (this.cachedUpstreamUsage && now - this.lastChecked < 60000) {
+      return this.cachedUpstreamUsage;
+    }
+
+    const sessionId = generateSessionId();
+    try {
+      const resp = await fetch(`${CONFIG.UPSTREAM_BASE_URL}/usage`, {
+        method: "GET",
+        headers: {
+          "Authorization": `Bearer ${apiKey}`,
+          "User-Agent": CONFIG.USER_AGENT,
+          "x-opencode-client": CONFIG.CLIENT,
+          "x-opencode-session": sessionId,
+          "x-opencode-request": crypto.randomUUID()
+        }
+      });
+
+      if (resp.ok) {
+        this.cachedUpstreamUsage = await resp.json();
+        this.lastChecked = now;
+        return this.cachedUpstreamUsage;
+      }
+    } catch {
+      // Endpoint may not be implemented upstream or require workspace ID
+    }
+
+    return null;
+  }
+
+  getStatus() {
+    return {
+      sessionRequests: this.sessionRequests,
+      sessionTokens: this.sessionTokens,
+      upstream: this.cachedUpstreamUsage
+    };
+  }
+}
+
+export const usageTracker = new UsageTracker();
diff --git a/skills/opencode-unlocked/SKILL.md b/skills/opencode-unlocked/SKILL.md
new file mode 100644
index 0000000..1ee824e
--- /dev/null
+++ b/skills/opencode-unlocked/SKILL.md
@@ -0,0 +1,29 @@
+---
+name: opencode-unlocked
+description: Guide and operational reference for using OpenCode Go models via the Opencode Unlocked bridge in OpenClaude.
+---
+
+# Opencode Unlocked Skill
+
+This skill explains how OpenClaude interacts with OpenCode Go models through the **Opencode Unlocked** plugin and local harness bridge.
+
+## Overview
+OpenCode Go hosts 36+ cutting-edge models behind strict harness fingerprinting. The **Opencode Unlocked** plugin runs a zero-dependency local bridge daemon on `http://127.0.0.1:8788/v1` that handles:
+1. **Authentic Harness Headers**: Injects exact 30-character OpenCode v2 session IDs (`ses_<timestamp_base62>`), client affinity, and harness metadata.
+2. **Transparent Protocol Translation**: Translates `/v1/chat/completions` into the OpenAI `/responses` API for models that require it (such as **`muse-spark-1.3-contributor`** and **`gpt-5.6-luna`**), while natively passing standard Chat Completions for models like **`space-bunny`**.
+3. **Model Prefix Stripping**: Normalizes prefix variations (`opencode-unlocked-*` -> `*`).
+
+## Primary Models
+- **`muse-spark-1.3-contributor`** (Default): High-capability reasoning and coding model. Requires `/responses` protocol; translated automatically by the bridge.
+- **`space-bunny`**: Fast, lightweight general and coding model. Native `/chat/completions`.
+- **`deepseek-v4-flash`**: High-throughput flash model.
+- **`glm-5.3`**: General language model.
+- **`gpt-5.6-luna`**: High-tier reasoning model.
+- **`kimi-k2.5`**: Long-context reasoning model.
+- **`minimax-01`**: Fast conversational model.
+- **`qwen3.5-max`**: Large general capability model.
+
+## Operational Commands
+- Check daemon health and usage: `/opencode-unlocked:status`
+- Manage models and keys: `/opencode-unlocked:unlocked`
+- CLI Management: `opencode-unlocked [start|stop|restart|status|models|test|set-key]`
diff --git a/test/gateway.test.mjs b/test/gateway.test.mjs
new file mode 100644
index 0000000..c697519
--- /dev/null
+++ b/test/gateway.test.mjs
@@ -0,0 +1,93 @@
+import assert from "node:assert/strict";
+import { createGatewayServer } from "../server/gateway.mjs";
+import { getApiKey } from "../server/config.mjs";
+
+console.log("=== Running Gateway Integration Tests ===");
+
+const TEST_PORT = 8799;
+const server = createGatewayServer();
+
+server.listen(TEST_PORT, "127.0.0.1", async () => {
+  const baseUrl = `http://127.0.0.1:${TEST_PORT}`;
+  const apiKey = getApiKey();
+
+  try {
+    // 1. Health check
+    const healthRes = await fetch(`${baseUrl}/health`);
+    assert.equal(healthRes.status, 200);
+    const healthData = await healthRes.json();
+    assert.equal(healthData.status, "ok");
+    console.log("✔ Health check passed");
+
+    // 2. Models list
+    const modelsRes = await fetch(`${baseUrl}/v1/models`);
+    assert.equal(modelsRes.status, 200);
+    const modelsData = await modelsRes.json();
+    assert.ok(Array.isArray(modelsData.data));
+    assert.ok(modelsData.data.length > 0);
+    console.log(`✔ Models list passed (${modelsData.data.length} models)`);
+
+    if (apiKey) {
+      // 3. Test space-bunny non-streaming
+      console.log("Testing space-bunny non-streaming...");
+      const spaceRes = await fetch(`${baseUrl}/v1/chat/completions`, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          model: "space-bunny",
+          messages: [{ role: "user", content: "Reply with exactly 'OK-SPACE'" }],
+          stream: false,
+          max_tokens: 20
+        })
+      });
+      assert.equal(spaceRes.status, 200, "space-bunny should return 200");
+      const spaceData = await spaceRes.json();
+      assert.ok(spaceData.choices?.[0]?.message?.content);
+      console.log(`✔ space-bunny non-streaming passed: "${spaceData.choices[0].message.content.trim()}"`);
+
+      // 4. Test muse-spark-1.3-contributor non-streaming (Responses API translation)
+      console.log("Testing muse-spark-1.3-contributor non-streaming (translated to /responses)...");
+      const museRes = await fetch(`${baseUrl}/v1/chat/completions`, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          model: "muse-spark-1.3-contributor",
+          messages: [{ role: "user", content: "Reply with exactly 'OK-MUSE'" }],
+          stream: false,
+          max_tokens: 1000
+        })
+      });
+      assert.equal(museRes.status, 200, "muse-spark-1.3-contributor should return 200");
+      const museData = await museRes.json();
+      assert.ok(museData.choices?.[0]?.message?.content);
+      console.log(`✔ muse-spark-1.3-contributor non-streaming passed: "${museData.choices[0].message.content.trim()}"`);
+
+      // 5. Test muse-spark-1.3-contributor streaming (SSE translation)
+      console.log("Testing muse-spark-1.3-contributor streaming (translated to SSE)...");
+      const streamRes = await fetch(`${baseUrl}/v1/chat/completions`, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          model: "muse-spark-1.3-contributor",
+          messages: [{ role: "user", content: "Reply with 'PING'" }],
+          stream: true,
+          max_tokens: 1000
+        })
+      });
+      assert.equal(streamRes.status, 200, "Streaming should return 200");
+      const streamText = await streamRes.text();
+      assert.ok(streamText.includes("data: "), "Stream must contain SSE data blocks");
+      assert.ok(streamText.includes("[DONE]"), "Stream must conclude with [DONE]");
+      console.log("✔ muse-spark-1.3-contributor streaming passed");
+    } else {
+      console.log("⚠ Skipping live API requests (no key configured)");
+    }
+
+    console.log("✔ All Gateway Integration Tests Passed!");
+  } catch (err) {
+    console.error("❌ Gateway integration test failed:", err);
+    process.exitCode = 1;
+  } finally {
+    server.close();
+  }
+});
diff --git a/test/live-models.test.mjs b/test/live-models.test.mjs
new file mode 100644
index 0000000..f02c822
--- /dev/null
+++ b/test/live-models.test.mjs
@@ -0,0 +1,79 @@
+import assert from "node:assert/strict";
+import { createGatewayServer } from "../server/gateway.mjs";
+import { getApiKey } from "../server/config.mjs";
+
+console.log("==================================================");
+console.log("   Opencode Unlocked: Live Models Verification    ");
+console.log("==================================================");
+
+const apiKey = getApiKey();
+if (!apiKey) {
+  console.error("❌ No API key available for live test.");
+  process.exit(1);
+}
+
+const server = createGatewayServer();
+const TEST_PORT = 8797;
+
+server.listen(TEST_PORT, "127.0.0.1", async () => {
+  const baseUrl = `http://127.0.0.1:${TEST_PORT}/v1`;
+
+  try {
+    // 1. Verify Default Model: Muse Spark 1.3 Contributor
+    console.log("\n[1/2] Verifying Target Default Model: Muse Spark 1.3 Contributor...");
+    const musePrompt = "What is 7 * 8? Reply with only the number.";
+    const museStart = Date.now();
+
+    const museRes = await fetch(`${baseUrl}/chat/completions`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({
+        model: "muse-spark-1.3-contributor",
+        messages: [{ role: "user", content: musePrompt }],
+        stream: false
+      })
+    });
+
+    assert.equal(museRes.status, 200, `Muse Spark HTTP status should be 200 (was ${museRes.status})`);
+    const museJson = await museRes.json();
+    const museReply = museJson.choices?.[0]?.message?.content?.trim();
+    const museDuration = ((Date.now() - museStart) / 1000).toFixed(2);
+    console.log(`✔ Muse Spark 1.3 Contributor responded in ${museDuration}s:`);
+    console.log(`  Output: "${museReply}"`);
+    console.log(`  Usage: ${JSON.stringify(museJson.usage)}`);
+    assert.ok(museReply && museReply.includes("56"), "Muse Spark should correctly calculate 7 * 8 = 56");
+
+    // 2. Verify Alternative Model: Space Bunny
+    console.log("\n[2/2] Verifying Target Model: Space Bunny...");
+    const spacePrompt = "What color is a ripe banana? Reply in one word.";
+    const spaceStart = Date.now();
+
+    const spaceRes = await fetch(`${baseUrl}/chat/completions`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({
+        model: "space-bunny",
+        messages: [{ role: "user", content: spacePrompt }],
+        stream: false
+      })
+    });
+
+    assert.equal(spaceRes.status, 200, `Space Bunny HTTP status should be 200 (was ${spaceRes.status})`);
+    const spaceJson = await spaceRes.json();
+    const spaceReply = spaceJson.choices?.[0]?.message?.content?.trim();
+    const spaceDuration = ((Date.now() - spaceStart) / 1000).toFixed(2);
+    console.log(`✔ Space Bunny responded in ${spaceDuration}s:`);
+    console.log(`  Output: "${spaceReply}"`);
+    console.log(`  Usage: ${JSON.stringify(spaceJson.usage)}`);
+    assert.ok(spaceReply && /yellow/i.test(spaceReply), "Space Bunny should correctly reply with yellow");
+
+    console.log("\n==================================================");
+    console.log("✔ LIVE VERIFICATION PASSED FOR BOTH MODELS!");
+    console.log("==================================================");
+  } catch (err) {
+    console.error("\n❌ Live model verification failed:", err);
+    process.exitCode = 1;
+  } finally {
+    server.close();
+  }
+});
diff --git a/test/session.test.mjs b/test/session.test.mjs
new file mode 100644
index 0000000..5e80fb2
--- /dev/null
+++ b/test/session.test.mjs
@@ -0,0 +1,32 @@
+import assert from "node:assert/strict";
+import { generateDescendingIdentifier, generateSessionId, isValidSessionId, sessionAffinity } from "../server/session.mjs";
+
+console.log("=== Running Session Tests ===");
+
+// 1. Validate session ID length and regex
+const id1 = generateSessionId();
+console.log(`Generated Session ID: ${id1}`);
+assert.equal(id1.length, 30, "Session ID must be exactly 30 characters");
+assert.equal(id1.startsWith("ses_"), true, "Session ID must start with 'ses_'");
+assert.equal(isValidSessionId(id1), true, "isValidSessionId must return true for generated ID");
+
+// 2. Validate descending time property
+const now = Date.now();
+const earlierId = generateDescendingIdentifier(now - 100000);
+const laterId = generateDescendingIdentifier(now);
+assert.ok(
+  earlierId > laterId,
+  "Earlier timestamp must have greater string value in descending sort order"
+);
+
+// 3. Validate affinity manager
+const clientA = "conv-123";
+const sidA1 = sessionAffinity.getOrCreate(clientA);
+const sidA2 = sessionAffinity.getOrCreate(clientA);
+assert.equal(sidA1, sidA2, "Same client identifier must receive identical session ID");
+
+const clientB = "conv-456";
+const sidB = sessionAffinity.getOrCreate(clientB);
+assert.notEqual(sidA1, sidB, "Different clients must receive distinct session IDs");
+
+console.log("✔ All Session Tests Passed!");
diff --git a/test/translator.test.mjs b/test/translator.test.mjs
new file mode 100644
index 0000000..aa18f42
--- /dev/null
+++ b/test/translator.test.mjs
@@ -0,0 +1,76 @@
+import assert from "node:assert/strict";
+import {
+  requiresResponsesApi,
+  chatCompletionsToResponsesBody,
+  responsesToChatCompletionsResponse,
+  translateResponsesEventToOpenAiChunk
+} from "../server/translator.mjs";
+
+console.log("=== Running Protocol Translator Tests ===");
+
+// 1. Verify requiresResponsesApi logic
+assert.equal(requiresResponsesApi("muse-spark-1.3-contributor"), true);
+assert.equal(requiresResponsesApi("opencode-unlocked-muse-spark-1.3-contributor"), true);
+assert.equal(requiresResponsesApi("gpt-5.6-luna"), true);
+assert.equal(requiresResponsesApi("space-bunny"), false);
+assert.equal(requiresResponsesApi("deepseek-v4-flash"), false);
+
+// 2. Chat Completions -> Responses payload conversion
+const openAiReq = {
+  model: "opencode-unlocked-muse-spark-1.3-contributor",
+  messages: [
+    { role: "system", content: "You are a helpful assistant." },
+    { role: "user", content: "Hello world" }
+  ],
+  stream: true,
+  temperature: 0.5,
+  max_tokens: 1000
+};
+
+const responsesPayload = chatCompletionsToResponsesBody(openAiReq);
+assert.equal(responsesPayload.model, "muse-spark-1.3-contributor");
+assert.equal(responsesPayload.stream, true);
+assert.equal(responsesPayload.temperature, 0.5);
+assert.equal(responsesPayload.max_output_tokens, 1000);
+assert.equal(responsesPayload.input.length, 2);
+assert.equal(responsesPayload.input[0].role, "system");
+assert.equal(responsesPayload.input[1].content, "Hello world");
+
+// 3. Responses API non-streaming JSON -> Chat Completions JSON
+const rawResponsesResp = {
+  id: "resp_1234567890",
+  model: "muse-spark-1.3-contributor",
+  output: [
+    {
+      type: "message",
+      role: "assistant",
+      content: [{ type: "output_text", text: "Hello from Responses API!" }]
+    }
+  ],
+  stop_reason: "stop",
+  usage: {
+    input_tokens: 15,
+    output_tokens: 8
+  }
+};
+
+const chatResp = responsesToChatCompletionsResponse(rawResponsesResp, "muse-spark-1.3-contributor");
+assert.equal(chatResp.id, "chatcmpl_1234567890");
+assert.equal(chatResp.choices[0].message.content, "Hello from Responses API!");
+assert.equal(chatResp.choices[0].finish_reason, "stop");
+assert.equal(chatResp.usage.prompt_tokens, 15);
+assert.equal(chatResp.usage.completion_tokens, 8);
+assert.equal(chatResp.usage.total_tokens, 23);
+
+// 4. SSE translation
+const state = { id: "", model: "muse-spark-1.3-contributor" };
+const textChunkRaw = `event: response.output_text.delta\ndata: {"type":"response.output_text.delta","delta":"Hello "}\n\n`;
+const textChunkOpenAi = translateResponsesEventToOpenAiChunk(textChunkRaw, state);
+assert.ok(textChunkOpenAi.startsWith("data: "));
+const parsedChunk = JSON.parse(textChunkOpenAi.replace("data: ", "").trim());
+assert.equal(parsedChunk.choices[0].delta.content, "Hello ");
+
+const doneChunk = translateResponsesEventToOpenAiChunk("data: [DONE]\n\n", state);
+assert.equal(doneChunk, "data: [DONE]\n\n");
+
+console.log("✔ All Protocol Translator Tests Passed!");