anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 02:05:26 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

MCP Servers are not available if config is per project #2678

New Issue
Open
opened 2026-02-16 17:36:48 -05:00 by yindo · 5 comments
yindo commented 2026-02-16 17:36:48 -05:00
Owner
Copy Link
Copy Source

Originally created by @xTamasu on GitHub (Nov 7, 2025).

Originally assigned to: @rekram1-node on GitHub.

Description

MCP Servers are not available/detected if you define a config.json at your project-level. They work fine if you move the same config file to ~/.config/opencode/opencode.json.

OpenCode version

1.0.39

Steps to reproduce

  1. Create config at ~/.config/opencode/opencode.json

Insert:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "mcp_everything": {
      "type": "local",
      "command": ["npx", "-y", "@modelcontextprotocol/server-everything"],
    },
  },
}
  1. Run ``opencode run "use the mcp_everything tool to add the number 3 and 4. You're not allowed to calculate by yourself." Alternatively open opencode and run /status.
  2. Check if mcp usage is shown.
  3. Move ~/.config/opencode/opencode.json to your project root folder (.opencode/opencode.json)
  4. Run ``opencode run "use the mcp_everything tool to add the number 3 and 4. You're not allowed to calculate by yourself." Alternatively open opencode and run /status.
  5. Check if mcp usage is shown.

Screenshot and/or share link

No response

Operating System

macOS 15.6.1

Terminal

vs-code, zsh

Originally created by @xTamasu on GitHub (Nov 7, 2025). Originally assigned to: @rekram1-node on GitHub. ### Description MCP Servers are not available/detected if you define a config.json at your project-level. They work fine if you move the same config file to ~/.config/opencode/opencode.json. ### OpenCode version 1.0.39 ### Steps to reproduce 1. Create config at ~/.config/opencode/opencode.json Insert: ``` { "$schema": "https://opencode.ai/config.json", "mcp": { "mcp_everything": { "type": "local", "command": ["npx", "-y", "@modelcontextprotocol/server-everything"], }, }, } ``` 2. Run ``opencode run "use the mcp_everything tool to add the number 3 and 4. You're not allowed to calculate by yourself." Alternatively open opencode and run /status. 3. Check if mcp usage is shown. 4. Move ~/.config/opencode/opencode.json to your project root folder (.opencode/opencode.json) 5. Run ``opencode run "use the mcp_everything tool to add the number 3 and 4. You're not allowed to calculate by yourself." Alternatively open opencode and run /status. 6. Check if mcp usage is shown. ### Screenshot and/or share link _No response_ ### Operating System macOS 15.6.1 ### Terminal vs-code, zsh
yindo added the bug label 2026-02-16 17:36:48 -05:00
yindo commented 2026-02-16 17:36:48 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Nov 7, 2025):

@xTamasu this is something we will fix, but we don't read in .opencode/opencode.json we only read it at the root of the project or globally, but this is a common confusion so we will fix that

@rekram1-node commented on GitHub (Nov 7, 2025): @xTamasu this is something we will fix, but we don't read in `.opencode/opencode.json` we only read it at the root of the project or globally, but this is a common confusion so we will fix that
yindo commented 2026-02-16 17:36:48 -05:00
Author
Owner
Copy Link
Copy Source

@xTamasu commented on GitHub (Nov 7, 2025):

@rekram1-node What exactly means "at the root of the project"? How should it be placed? Its enough for me if I can at least somehow work non-global.

@xTamasu commented on GitHub (Nov 7, 2025): @rekram1-node What exactly means "at the root of the project"? How should it be placed? Its enough for me if I can at least somehow work non-global.
yindo commented 2026-02-16 17:36:48 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Nov 7, 2025):

@xTamasu any opencode.json or opencode.jsonc in your current working directory or in a parent directory should be picked up.

Now if this is a "project" aka it is a cloned git repo for example, then the root of your project is where the README.md and .gitignore files typically are.

I will also note that in the next release we will read from .opencode/ directories too tho

fix

@rekram1-node commented on GitHub (Nov 7, 2025): @xTamasu any opencode.json or opencode.jsonc in your current working directory or in a parent directory should be picked up. Now if this is a "project" aka it is a cloned git repo for example, then the root of your project is where the README.md and .gitignore files typically are. I will also note that in the next release we will read from .opencode/ directories too tho [fix](https://github.com/sst/opencode/commit/8cef7940fed5f12cee55a54b7c4c9c0ff1b1a322)
yindo commented 2026-02-16 17:36:48 -05:00
Author
Owner
Copy Link
Copy Source

@An0nX commented on GitHub (Dec 30, 2025):

Image Image

got same ussue. while trying to use opencode mcp add opencode says that mcp was added successful, but nothing changed in the config file

{
  "$schema": "https://opencode.ai/config.json",

  // Теперь в global лежит только core (остальное уехало в skills).
  "instructions": ["./instructions/global/*.md"],

  "small_model": "openrouter/mimo-flash",

  "formatter": {
    "ruff": {
      "command": ["ruff", "format", "$FILE"],
      "extensions": [".py", ".pyi"],
      "environment": {
        "RUFF_NO_CACHE": "true",
        "RUFF_FORMAT_QUOTE_STYLE": "double",
        "RUFF_FORMAT_INDENT_STYLE": "space",
        "RUFF_FORMAT_SKIP_MAGIC_TRAILING_COMMA": "false",
        "RUFF_FORMAT_LINE_ENDING": "lf"
      }
    },

    // 3.4 — добавляем форматтер для фронта (как минимум JS/TS/CSS/HTML/JSON/MD).
    // Команду можно заменить на локально установленный prettier (например, ["pnpm","prettier",...])
    "prettier": {
      "command": ["npx", "prettier", "--write", "$FILE"],
      "extensions": [
        ".js",
        ".jsx",
        ".ts",
        ".tsx",
        ".css",
        ".scss",
        ".html",
        ".json",
        ".md",
        ".yaml",
        ".yml"
      ]
    }
  },

  "provider": {
    "openrouter": {
      "models": {
        "mimo-flash": {
          "id": "xiaomi/mimo-v2-flash:free",
          "name": "Xiaomi MiMo V2 Flash"
        },
        "gemini-flash": {
          "id": "google/gemini-2.0-flash-exp:free",
          "name": "Google Gemini 2.0 Flash Experimental"
        },
        "devstral": {
          "id": "mistralai/devstral-2512:free",
          "name": "Mistral Devstral 2"
        },
        "deepseek-r1t2": {
          "id": "tngtech/deepseek-r1t2-chimera:free",
          "name": "DeepSeek R1T2 Chimera"
        },
        "r1t-chimera": {
          "id": "tngtech/tng-r1t-chimera:free",
          "name": "R1T Chimera"
        }
      }
    }
  },

  "agent": {
    "build": {
      "mode": "primary",
      "model": "openrouter/mimo-flash",
      "description": "Primary Orchestrator / Integrator.\n\nWhat it does:\n- Owns end-to-end delivery: analyze → plan/delegate → implement → verify → document.\n- Loads the correct stack skill(s) (stack-django/stack-python/stack-frontend) before changing code.\n- Delegates specialized work to subagents (plan/docker/cicd/security/qa-tests/reviewer/docs/version-scout).\n- Uses MCP memory to persist project invariants (commands, ports, conventions, architectural decisions) across sessions.\n\nWhen to use:\n- Default agent for any coding task touching multiple files or requiring coordination.\n\nOutputs:\n- Concrete code/config changes + short verification steps + what to run locally/CI.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/build.md}",
      "tools": {
        "read": true,
        "write": true,
        "edit": true,
        "bash": true,
        "skill": true,
        "mcp": true
      }
    },

    "plan": {
      "mode": "primary",
      "model": "openrouter/mimo-flash",
      "description": "Architecture & Execution Planner.\n\nWhat it does:\n- Produces dependency-aware plans, risk analysis, and handoff instructions.\n- Loads relevant stack skill(s) before planning so the plan matches the project constraints.\n- Uses MCP memory to reuse prior architectural decisions and to persist new decisions.\n\nWhen to use:\n- Complex features, refactors, multi-service changes, unclear boundaries.\n\nOutputs:\n- Structured Markdown plan with phases, DoD, dependency graph, and risks.",
      "temperature": 0.2,
      "prompt": "{file:./instructions/inject/plan.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "qa-tests": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "QA Strategy & Test Design (pre-implementation).\n\nWhat it does:\n- Builds regression-proof test matrix (happy path, edge cases, failures, observability).\n- Enforces 'test first for bugfix': test must fail before fix and pass after.\n- Uses memory to persist agreed test commands and key scenarios.\n\nWhen to use:\n- Any non-trivial logic change; before implementation to define DoD for tests.\n\nOutputs:\n- Markdown test matrix + execution plan (without TODO/placeholders in executable snippets).",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/qa-tests.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "docker": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Containerization & Runtime Engineer.\n\nWhat it does:\n- Authors/reviews Dockerfile/docker-compose for correctness, security, caching, non-root, signals.\n- Follows your policy: prefer python:slim / python:alpine (no digest pinning).\n- Uses memory to persist build/run commands, ports, and container conventions.\n\nWhen to use:\n- Any Dockerfile/compose changes or runtime env questions.\n\nOutputs:\n- Full updated Dockerfile/compose + rationale + how to run/verify.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/docker.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "cicd": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "CI/CD Pipeline Engineer.\n\nWhat it does:\n- Designs fast, deterministic, secure pipelines (GitHub Actions by default).\n- Minimizes privileges, maximizes caching.\n- Uses memory to persist canonical CI commands and pipeline conventions.\n\nWhen to use:\n- Any workflow/pipeline changes.\n\nOutputs:\n- Ready-to-apply YAML workflows (no TODO placeholders; unknowns marked as UNKNOWN with a discovery plan).",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/cicd.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "reviewer": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Adversarial Code Reviewer (read-only).\n\nWhat it does:\n- Reviews diffs for correctness, performance, readability.\n- Also checks: migrations/DB safety, backward compatibility, secrets/logging/PII.\n- Uses memory to keep project-specific review checklist (if stable).\n\nWhen to use:\n- Always after implementation before shipping.\n\nOutputs:\n- Actionable findings with file:line, issue, fix, why.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/reviewer.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "security": {
      "mode": "subagent",
      "model": "openrouter/deepseek-r1t2",
      "description": "Security / Threat Modeling (white hat).\n\nWhat it does:\n- Threat models and audits for common vuln classes (SQLi/XSS/SSRF/IDOR/supply-chain).\n- Produces exploit scenario + exact remediation.\n- Uses memory to persist known attack surfaces and security decisions.\n\nWhen to use:\n- Auth, secrets, input handling, network boundaries, permission systems.\n\nOutputs:\n- Findings categorized CRITICAL/HIGH/MEDIUM + verification checklist.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/security.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "ideator": {
      "mode": "subagent",
      "model": "openrouter/r1t-chimera",
      "description": "Product / Solution Ideation (Tree of Thoughts).\n\nWhat it does:\n- Generates MVP/Strategic/Moonshot concept branches.\n- Helps when requirements unclear or you want options.\n- Can store stable product constraints in memory if useful.\n\nWhen to use:\n- Early discovery, brainstorming, when user wants multiple approaches.\n\nOutputs:\n- 3 concepts with user story, feasibility, wow-factor.",
      "temperature": 0.8,
      "prompt": "{file:./instructions/inject/ideator.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "docs": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Documentation / Developer Advocate.\n\nWhat it does:\n- Updates README and API docs so docs stay source-of-truth.\n- Ensures code blocks are copy/paste runnable.\n- Uses memory to persist canonical 'how to run/test' snippets.\n\nWhen to use:\n- Always when APIs change, new commands appear, or setup changes.\n\nOutputs:\n- List of files to update + full Markdown content.",
      "temperature": 0.2,
      "prompt": "{file:./instructions/inject/docs.md}",
      "tools": {
        "read": true,
        "write": true,
        "edit": true,
        "skill": true,
        "mcp": true
      }
    },

    "version-scout": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Dependency / Version Researcher (latest stable).\n\nWhat it does:\n- Uses webfetch to find latest stable releases for libraries/tools/images.\n- Never guesses versions. If unknown → UNKNOWN.\n- Also checks if project is on an older version and calls it out.\n- Uses memory to store 'last verified version + date' (but re-verifies if 'latest' is requested).\n\nWhen to use:\n- Poetry deps, tool upgrades, when you suspect repo is behind.\n\nOutputs:\n- Markdown table: component, latest stable version, release date, source link, notes.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/version-scout.md}",
      "tools": {
        "read": true,
        "webfetch": true,
        "skill": true,
        "mcp": true
      }
    }
  },

  "permission": {
    "read": "allow",
    "edit": "allow",
    "write": "allow",
    "webfetch": "allow",

    "skill": { "*": "allow" },

    // 3.3 — фиксируем явно, чтобы поведение было читаемо в конфиге.
    "doom_loop": "ask",
    "external_directory": "ask",

    // 3.3 — делаем ask на явно опасные команды, остальное оставляем allow.
    "bash": {
      "git commit*": "allow",
      "git push": "ask",

      "rm *": "ask",
      "sudo *": "ask",
      "git reset *": "ask",
      "git clean *": "ask",
      "docker system prune*": "ask",

      "*": "allow"
    },

    "mcp": {
      "context7": {
        "type": "local",
        "command": ["npx", "-y", "@upstash/context7-mcp@latest"],
        "enabled": true,
        "timeout": 60000
      },
      "sequential-thinking": {
        "type": "local",
        "command": ["npx", "-y", "@modelcontextprotocol/server-sequential-thinking"],
        "enabled": true,
        "timeout": 60000
      },
      "memory": {
        "type": "local",
        "command": ["npx", "-y", "@modelcontextprotocol/server-memory"],
        "enabled": true,
        "timeout": 60000
      }
    }
  }
}
@An0nX commented on GitHub (Dec 30, 2025): <img width="1840" height="955" alt="Image" src="https://github.com/user-attachments/assets/1f2194de-1f1c-4e9a-a607-13899626c516" /> <img width="538" height="261" alt="Image" src="https://github.com/user-attachments/assets/00bee7ff-e295-450e-8ddf-0931f0c41cd9" /> got same ussue. while trying to use `opencode mcp add` opencode says that mcp was added successful, but nothing changed in the config file ``` { "$schema": "https://opencode.ai/config.json", // Теперь в global лежит только core (остальное уехало в skills). "instructions": ["./instructions/global/*.md"], "small_model": "openrouter/mimo-flash", "formatter": { "ruff": { "command": ["ruff", "format", "$FILE"], "extensions": [".py", ".pyi"], "environment": { "RUFF_NO_CACHE": "true", "RUFF_FORMAT_QUOTE_STYLE": "double", "RUFF_FORMAT_INDENT_STYLE": "space", "RUFF_FORMAT_SKIP_MAGIC_TRAILING_COMMA": "false", "RUFF_FORMAT_LINE_ENDING": "lf" } }, // 3.4 — добавляем форматтер для фронта (как минимум JS/TS/CSS/HTML/JSON/MD). // Команду можно заменить на локально установленный prettier (например, ["pnpm","prettier",...]) "prettier": { "command": ["npx", "prettier", "--write", "$FILE"], "extensions": [ ".js", ".jsx", ".ts", ".tsx", ".css", ".scss", ".html", ".json", ".md", ".yaml", ".yml" ] } }, "provider": { "openrouter": { "models": { "mimo-flash": { "id": "xiaomi/mimo-v2-flash:free", "name": "Xiaomi MiMo V2 Flash" }, "gemini-flash": { "id": "google/gemini-2.0-flash-exp:free", "name": "Google Gemini 2.0 Flash Experimental" }, "devstral": { "id": "mistralai/devstral-2512:free", "name": "Mistral Devstral 2" }, "deepseek-r1t2": { "id": "tngtech/deepseek-r1t2-chimera:free", "name": "DeepSeek R1T2 Chimera" }, "r1t-chimera": { "id": "tngtech/tng-r1t-chimera:free", "name": "R1T Chimera" } } } }, "agent": { "build": { "mode": "primary", "model": "openrouter/mimo-flash", "description": "Primary Orchestrator / Integrator.\n\nWhat it does:\n- Owns end-to-end delivery: analyze → plan/delegate → implement → verify → document.\n- Loads the correct stack skill(s) (stack-django/stack-python/stack-frontend) before changing code.\n- Delegates specialized work to subagents (plan/docker/cicd/security/qa-tests/reviewer/docs/version-scout).\n- Uses MCP memory to persist project invariants (commands, ports, conventions, architectural decisions) across sessions.\n\nWhen to use:\n- Default agent for any coding task touching multiple files or requiring coordination.\n\nOutputs:\n- Concrete code/config changes + short verification steps + what to run locally/CI.", "temperature": 0.1, "prompt": "{file:./instructions/inject/build.md}", "tools": { "read": true, "write": true, "edit": true, "bash": true, "skill": true, "mcp": true } }, "plan": { "mode": "primary", "model": "openrouter/mimo-flash", "description": "Architecture & Execution Planner.\n\nWhat it does:\n- Produces dependency-aware plans, risk analysis, and handoff instructions.\n- Loads relevant stack skill(s) before planning so the plan matches the project constraints.\n- Uses MCP memory to reuse prior architectural decisions and to persist new decisions.\n\nWhen to use:\n- Complex features, refactors, multi-service changes, unclear boundaries.\n\nOutputs:\n- Structured Markdown plan with phases, DoD, dependency graph, and risks.", "temperature": 0.2, "prompt": "{file:./instructions/inject/plan.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "qa-tests": { "mode": "subagent", "model": "openrouter/mimo-flash", "description": "QA Strategy & Test Design (pre-implementation).\n\nWhat it does:\n- Builds regression-proof test matrix (happy path, edge cases, failures, observability).\n- Enforces 'test first for bugfix': test must fail before fix and pass after.\n- Uses memory to persist agreed test commands and key scenarios.\n\nWhen to use:\n- Any non-trivial logic change; before implementation to define DoD for tests.\n\nOutputs:\n- Markdown test matrix + execution plan (without TODO/placeholders in executable snippets).", "temperature": 0.1, "prompt": "{file:./instructions/inject/qa-tests.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "docker": { "mode": "subagent", "model": "openrouter/mimo-flash", "description": "Containerization & Runtime Engineer.\n\nWhat it does:\n- Authors/reviews Dockerfile/docker-compose for correctness, security, caching, non-root, signals.\n- Follows your policy: prefer python:slim / python:alpine (no digest pinning).\n- Uses memory to persist build/run commands, ports, and container conventions.\n\nWhen to use:\n- Any Dockerfile/compose changes or runtime env questions.\n\nOutputs:\n- Full updated Dockerfile/compose + rationale + how to run/verify.", "temperature": 0.1, "prompt": "{file:./instructions/inject/docker.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "cicd": { "mode": "subagent", "model": "openrouter/mimo-flash", "description": "CI/CD Pipeline Engineer.\n\nWhat it does:\n- Designs fast, deterministic, secure pipelines (GitHub Actions by default).\n- Minimizes privileges, maximizes caching.\n- Uses memory to persist canonical CI commands and pipeline conventions.\n\nWhen to use:\n- Any workflow/pipeline changes.\n\nOutputs:\n- Ready-to-apply YAML workflows (no TODO placeholders; unknowns marked as UNKNOWN with a discovery plan).", "temperature": 0.1, "prompt": "{file:./instructions/inject/cicd.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "reviewer": { "mode": "subagent", "model": "openrouter/mimo-flash", "description": "Adversarial Code Reviewer (read-only).\n\nWhat it does:\n- Reviews diffs for correctness, performance, readability.\n- Also checks: migrations/DB safety, backward compatibility, secrets/logging/PII.\n- Uses memory to keep project-specific review checklist (if stable).\n\nWhen to use:\n- Always after implementation before shipping.\n\nOutputs:\n- Actionable findings with file:line, issue, fix, why.", "temperature": 0.1, "prompt": "{file:./instructions/inject/reviewer.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "security": { "mode": "subagent", "model": "openrouter/deepseek-r1t2", "description": "Security / Threat Modeling (white hat).\n\nWhat it does:\n- Threat models and audits for common vuln classes (SQLi/XSS/SSRF/IDOR/supply-chain).\n- Produces exploit scenario + exact remediation.\n- Uses memory to persist known attack surfaces and security decisions.\n\nWhen to use:\n- Auth, secrets, input handling, network boundaries, permission systems.\n\nOutputs:\n- Findings categorized CRITICAL/HIGH/MEDIUM + verification checklist.", "temperature": 0.1, "prompt": "{file:./instructions/inject/security.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "ideator": { "mode": "subagent", "model": "openrouter/r1t-chimera", "description": "Product / Solution Ideation (Tree of Thoughts).\n\nWhat it does:\n- Generates MVP/Strategic/Moonshot concept branches.\n- Helps when requirements unclear or you want options.\n- Can store stable product constraints in memory if useful.\n\nWhen to use:\n- Early discovery, brainstorming, when user wants multiple approaches.\n\nOutputs:\n- 3 concepts with user story, feasibility, wow-factor.", "temperature": 0.8, "prompt": "{file:./instructions/inject/ideator.md}", "tools": { "read": true, "skill": true, "mcp": true } }, "docs": { "mode": "subagent", "model": "openrouter/mimo-flash", "description": "Documentation / Developer Advocate.\n\nWhat it does:\n- Updates README and API docs so docs stay source-of-truth.\n- Ensures code blocks are copy/paste runnable.\n- Uses memory to persist canonical 'how to run/test' snippets.\n\nWhen to use:\n- Always when APIs change, new commands appear, or setup changes.\n\nOutputs:\n- List of files to update + full Markdown content.", "temperature": 0.2, "prompt": "{file:./instructions/inject/docs.md}", "tools": { "read": true, "write": true, "edit": true, "skill": true, "mcp": true } }, "version-scout": { "mode": "subagent", "model": "openrouter/mimo-flash", "description": "Dependency / Version Researcher (latest stable).\n\nWhat it does:\n- Uses webfetch to find latest stable releases for libraries/tools/images.\n- Never guesses versions. If unknown → UNKNOWN.\n- Also checks if project is on an older version and calls it out.\n- Uses memory to store 'last verified version + date' (but re-verifies if 'latest' is requested).\n\nWhen to use:\n- Poetry deps, tool upgrades, when you suspect repo is behind.\n\nOutputs:\n- Markdown table: component, latest stable version, release date, source link, notes.", "temperature": 0.1, "prompt": "{file:./instructions/inject/version-scout.md}", "tools": { "read": true, "webfetch": true, "skill": true, "mcp": true } } }, "permission": { "read": "allow", "edit": "allow", "write": "allow", "webfetch": "allow", "skill": { "*": "allow" }, // 3.3 — фиксируем явно, чтобы поведение было читаемо в конфиге. "doom_loop": "ask", "external_directory": "ask", // 3.3 — делаем ask на явно опасные команды, остальное оставляем allow. "bash": { "git commit*": "allow", "git push": "ask", "rm *": "ask", "sudo *": "ask", "git reset *": "ask", "git clean *": "ask", "docker system prune*": "ask", "*": "allow" }, "mcp": { "context7": { "type": "local", "command": ["npx", "-y", "@upstash/context7-mcp@latest"], "enabled": true, "timeout": 60000 }, "sequential-thinking": { "type": "local", "command": ["npx", "-y", "@modelcontextprotocol/server-sequential-thinking"], "enabled": true, "timeout": 60000 }, "memory": { "type": "local", "command": ["npx", "-y", "@modelcontextprotocol/server-memory"], "enabled": true, "timeout": 60000 } } } } ```
yindo commented 2026-02-16 17:36:48 -05:00
Author
Owner
Copy Link
Copy Source

@yaleh commented on GitHub (Jan 4, 2026):

Image Image
got same ussue. while trying to use opencode mcp add opencode says that mcp was added successful, but nothing changed in the config file

{
  "$schema": "https://opencode.ai/config.json",

  // Теперь в global лежит только core (остальное уехало в skills).
  "instructions": ["./instructions/global/*.md"],

  "small_model": "openrouter/mimo-flash",

  "formatter": {
    "ruff": {
      "command": ["ruff", "format", "$FILE"],
      "extensions": [".py", ".pyi"],
      "environment": {
        "RUFF_NO_CACHE": "true",
        "RUFF_FORMAT_QUOTE_STYLE": "double",
        "RUFF_FORMAT_INDENT_STYLE": "space",
        "RUFF_FORMAT_SKIP_MAGIC_TRAILING_COMMA": "false",
        "RUFF_FORMAT_LINE_ENDING": "lf"
      }
    },

    // 3.4 — добавляем форматтер для фронта (как минимум JS/TS/CSS/HTML/JSON/MD).
    // Команду можно заменить на локально установленный prettier (например, ["pnpm","prettier",...])
    "prettier": {
      "command": ["npx", "prettier", "--write", "$FILE"],
      "extensions": [
        ".js",
        ".jsx",
        ".ts",
        ".tsx",
        ".css",
        ".scss",
        ".html",
        ".json",
        ".md",
        ".yaml",
        ".yml"
      ]
    }
  },

  "provider": {
    "openrouter": {
      "models": {
        "mimo-flash": {
          "id": "xiaomi/mimo-v2-flash:free",
          "name": "Xiaomi MiMo V2 Flash"
        },
        "gemini-flash": {
          "id": "google/gemini-2.0-flash-exp:free",
          "name": "Google Gemini 2.0 Flash Experimental"
        },
        "devstral": {
          "id": "mistralai/devstral-2512:free",
          "name": "Mistral Devstral 2"
        },
        "deepseek-r1t2": {
          "id": "tngtech/deepseek-r1t2-chimera:free",
          "name": "DeepSeek R1T2 Chimera"
        },
        "r1t-chimera": {
          "id": "tngtech/tng-r1t-chimera:free",
          "name": "R1T Chimera"
        }
      }
    }
  },

  "agent": {
    "build": {
      "mode": "primary",
      "model": "openrouter/mimo-flash",
      "description": "Primary Orchestrator / Integrator.\n\nWhat it does:\n- Owns end-to-end delivery: analyze → plan/delegate → implement → verify → document.\n- Loads the correct stack skill(s) (stack-django/stack-python/stack-frontend) before changing code.\n- Delegates specialized work to subagents (plan/docker/cicd/security/qa-tests/reviewer/docs/version-scout).\n- Uses MCP memory to persist project invariants (commands, ports, conventions, architectural decisions) across sessions.\n\nWhen to use:\n- Default agent for any coding task touching multiple files or requiring coordination.\n\nOutputs:\n- Concrete code/config changes + short verification steps + what to run locally/CI.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/build.md}",
      "tools": {
        "read": true,
        "write": true,
        "edit": true,
        "bash": true,
        "skill": true,
        "mcp": true
      }
    },

    "plan": {
      "mode": "primary",
      "model": "openrouter/mimo-flash",
      "description": "Architecture & Execution Planner.\n\nWhat it does:\n- Produces dependency-aware plans, risk analysis, and handoff instructions.\n- Loads relevant stack skill(s) before planning so the plan matches the project constraints.\n- Uses MCP memory to reuse prior architectural decisions and to persist new decisions.\n\nWhen to use:\n- Complex features, refactors, multi-service changes, unclear boundaries.\n\nOutputs:\n- Structured Markdown plan with phases, DoD, dependency graph, and risks.",
      "temperature": 0.2,
      "prompt": "{file:./instructions/inject/plan.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "qa-tests": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "QA Strategy & Test Design (pre-implementation).\n\nWhat it does:\n- Builds regression-proof test matrix (happy path, edge cases, failures, observability).\n- Enforces 'test first for bugfix': test must fail before fix and pass after.\n- Uses memory to persist agreed test commands and key scenarios.\n\nWhen to use:\n- Any non-trivial logic change; before implementation to define DoD for tests.\n\nOutputs:\n- Markdown test matrix + execution plan (without TODO/placeholders in executable snippets).",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/qa-tests.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "docker": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Containerization & Runtime Engineer.\n\nWhat it does:\n- Authors/reviews Dockerfile/docker-compose for correctness, security, caching, non-root, signals.\n- Follows your policy: prefer python:slim / python:alpine (no digest pinning).\n- Uses memory to persist build/run commands, ports, and container conventions.\n\nWhen to use:\n- Any Dockerfile/compose changes or runtime env questions.\n\nOutputs:\n- Full updated Dockerfile/compose + rationale + how to run/verify.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/docker.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "cicd": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "CI/CD Pipeline Engineer.\n\nWhat it does:\n- Designs fast, deterministic, secure pipelines (GitHub Actions by default).\n- Minimizes privileges, maximizes caching.\n- Uses memory to persist canonical CI commands and pipeline conventions.\n\nWhen to use:\n- Any workflow/pipeline changes.\n\nOutputs:\n- Ready-to-apply YAML workflows (no TODO placeholders; unknowns marked as UNKNOWN with a discovery plan).",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/cicd.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "reviewer": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Adversarial Code Reviewer (read-only).\n\nWhat it does:\n- Reviews diffs for correctness, performance, readability.\n- Also checks: migrations/DB safety, backward compatibility, secrets/logging/PII.\n- Uses memory to keep project-specific review checklist (if stable).\n\nWhen to use:\n- Always after implementation before shipping.\n\nOutputs:\n- Actionable findings with file:line, issue, fix, why.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/reviewer.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "security": {
      "mode": "subagent",
      "model": "openrouter/deepseek-r1t2",
      "description": "Security / Threat Modeling (white hat).\n\nWhat it does:\n- Threat models and audits for common vuln classes (SQLi/XSS/SSRF/IDOR/supply-chain).\n- Produces exploit scenario + exact remediation.\n- Uses memory to persist known attack surfaces and security decisions.\n\nWhen to use:\n- Auth, secrets, input handling, network boundaries, permission systems.\n\nOutputs:\n- Findings categorized CRITICAL/HIGH/MEDIUM + verification checklist.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/security.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "ideator": {
      "mode": "subagent",
      "model": "openrouter/r1t-chimera",
      "description": "Product / Solution Ideation (Tree of Thoughts).\n\nWhat it does:\n- Generates MVP/Strategic/Moonshot concept branches.\n- Helps when requirements unclear or you want options.\n- Can store stable product constraints in memory if useful.\n\nWhen to use:\n- Early discovery, brainstorming, when user wants multiple approaches.\n\nOutputs:\n- 3 concepts with user story, feasibility, wow-factor.",
      "temperature": 0.8,
      "prompt": "{file:./instructions/inject/ideator.md}",
      "tools": {
        "read": true,
        "skill": true,
        "mcp": true
      }
    },

    "docs": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Documentation / Developer Advocate.\n\nWhat it does:\n- Updates README and API docs so docs stay source-of-truth.\n- Ensures code blocks are copy/paste runnable.\n- Uses memory to persist canonical 'how to run/test' snippets.\n\nWhen to use:\n- Always when APIs change, new commands appear, or setup changes.\n\nOutputs:\n- List of files to update + full Markdown content.",
      "temperature": 0.2,
      "prompt": "{file:./instructions/inject/docs.md}",
      "tools": {
        "read": true,
        "write": true,
        "edit": true,
        "skill": true,
        "mcp": true
      }
    },

    "version-scout": {
      "mode": "subagent",
      "model": "openrouter/mimo-flash",
      "description": "Dependency / Version Researcher (latest stable).\n\nWhat it does:\n- Uses webfetch to find latest stable releases for libraries/tools/images.\n- Never guesses versions. If unknown → UNKNOWN.\n- Also checks if project is on an older version and calls it out.\n- Uses memory to store 'last verified version + date' (but re-verifies if 'latest' is requested).\n\nWhen to use:\n- Poetry deps, tool upgrades, when you suspect repo is behind.\n\nOutputs:\n- Markdown table: component, latest stable version, release date, source link, notes.",
      "temperature": 0.1,
      "prompt": "{file:./instructions/inject/version-scout.md}",
      "tools": {
        "read": true,
        "webfetch": true,
        "skill": true,
        "mcp": true
      }
    }
  },

  "permission": {
    "read": "allow",
    "edit": "allow",
    "write": "allow",
    "webfetch": "allow",

    "skill": { "*": "allow" },

    // 3.3 — фиксируем явно, чтобы поведение было читаемо в конфиге.
    "doom_loop": "ask",
    "external_directory": "ask",

    // 3.3 — делаем ask на явно опасные команды, остальное оставляем allow (как у тебя было).
    "bash": {
      "git commit*": "allow",
      "git push": "ask",

      "rm *": "ask",
      "sudo *": "ask",
      "git reset *": "ask",
      "git clean *": "ask",
      "docker system prune*": "ask",

      "*": "allow"
    },

    "mcp": {
      "context7": {
        "type": "local",
        "command": ["npx", "-y", "@upstash/context7-mcp@latest"],
        "enabled": true,
        "timeout": 60000
      },
      "sequential-thinking": {
        "type": "local",
        "command": ["npx", "-y", "@modelcontextprotocol/server-sequential-thinking"],
        "enabled": true,
        "timeout": 60000
      },
      "memory": {
        "type": "local",
        "command": ["npx", "-y", "@modelcontextprotocol/server-memory"],
        "enabled": true,
        "timeout": 60000
      }
    }
  }
}

+1

@yaleh commented on GitHub (Jan 4, 2026): > <img alt="Image" width="1840" height="955" src="https://private-user-images.githubusercontent.com/80145620/530985260-1f2194de-1f1c-4e9a-a607-13899626c516.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.60yz0JpSD5NGkTDGp1bC0_fT9E9qx8mmBR8D2D74OyU"> <img alt="Image" width="538" height="261" src="https://private-user-images.githubusercontent.com/80145620/530985515-00bee7ff-e295-450e-8ddf-0931f0c41cd9.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.C4W6d_GhrJoy_9OOi9OkDuwJfpztmj9SeTdTkOg0JHU"> > got same ussue. while trying to use `opencode mcp add` opencode says that mcp was added successful, but nothing changed in the config file > > ``` > { > "$schema": "https://opencode.ai/config.json", > > // Теперь в global лежит только core (остальное уехало в skills). > "instructions": ["./instructions/global/*.md"], > > "small_model": "openrouter/mimo-flash", > > "formatter": { > "ruff": { > "command": ["ruff", "format", "$FILE"], > "extensions": [".py", ".pyi"], > "environment": { > "RUFF_NO_CACHE": "true", > "RUFF_FORMAT_QUOTE_STYLE": "double", > "RUFF_FORMAT_INDENT_STYLE": "space", > "RUFF_FORMAT_SKIP_MAGIC_TRAILING_COMMA": "false", > "RUFF_FORMAT_LINE_ENDING": "lf" > } > }, > > // 3.4 — добавляем форматтер для фронта (как минимум JS/TS/CSS/HTML/JSON/MD). > // Команду можно заменить на локально установленный prettier (например, ["pnpm","prettier",...]) > "prettier": { > "command": ["npx", "prettier", "--write", "$FILE"], > "extensions": [ > ".js", > ".jsx", > ".ts", > ".tsx", > ".css", > ".scss", > ".html", > ".json", > ".md", > ".yaml", > ".yml" > ] > } > }, > > "provider": { > "openrouter": { > "models": { > "mimo-flash": { > "id": "xiaomi/mimo-v2-flash:free", > "name": "Xiaomi MiMo V2 Flash" > }, > "gemini-flash": { > "id": "google/gemini-2.0-flash-exp:free", > "name": "Google Gemini 2.0 Flash Experimental" > }, > "devstral": { > "id": "mistralai/devstral-2512:free", > "name": "Mistral Devstral 2" > }, > "deepseek-r1t2": { > "id": "tngtech/deepseek-r1t2-chimera:free", > "name": "DeepSeek R1T2 Chimera" > }, > "r1t-chimera": { > "id": "tngtech/tng-r1t-chimera:free", > "name": "R1T Chimera" > } > } > } > }, > > "agent": { > "build": { > "mode": "primary", > "model": "openrouter/mimo-flash", > "description": "Primary Orchestrator / Integrator.\n\nWhat it does:\n- Owns end-to-end delivery: analyze → plan/delegate → implement → verify → document.\n- Loads the correct stack skill(s) (stack-django/stack-python/stack-frontend) before changing code.\n- Delegates specialized work to subagents (plan/docker/cicd/security/qa-tests/reviewer/docs/version-scout).\n- Uses MCP memory to persist project invariants (commands, ports, conventions, architectural decisions) across sessions.\n\nWhen to use:\n- Default agent for any coding task touching multiple files or requiring coordination.\n\nOutputs:\n- Concrete code/config changes + short verification steps + what to run locally/CI.", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/build.md}", > "tools": { > "read": true, > "write": true, > "edit": true, > "bash": true, > "skill": true, > "mcp": true > } > }, > > "plan": { > "mode": "primary", > "model": "openrouter/mimo-flash", > "description": "Architecture & Execution Planner.\n\nWhat it does:\n- Produces dependency-aware plans, risk analysis, and handoff instructions.\n- Loads relevant stack skill(s) before planning so the plan matches the project constraints.\n- Uses MCP memory to reuse prior architectural decisions and to persist new decisions.\n\nWhen to use:\n- Complex features, refactors, multi-service changes, unclear boundaries.\n\nOutputs:\n- Structured Markdown plan with phases, DoD, dependency graph, and risks.", > "temperature": 0.2, > "prompt": "{file:./instructions/inject/plan.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "qa-tests": { > "mode": "subagent", > "model": "openrouter/mimo-flash", > "description": "QA Strategy & Test Design (pre-implementation).\n\nWhat it does:\n- Builds regression-proof test matrix (happy path, edge cases, failures, observability).\n- Enforces 'test first for bugfix': test must fail before fix and pass after.\n- Uses memory to persist agreed test commands and key scenarios.\n\nWhen to use:\n- Any non-trivial logic change; before implementation to define DoD for tests.\n\nOutputs:\n- Markdown test matrix + execution plan (without TODO/placeholders in executable snippets).", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/qa-tests.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "docker": { > "mode": "subagent", > "model": "openrouter/mimo-flash", > "description": "Containerization & Runtime Engineer.\n\nWhat it does:\n- Authors/reviews Dockerfile/docker-compose for correctness, security, caching, non-root, signals.\n- Follows your policy: prefer python:slim / python:alpine (no digest pinning).\n- Uses memory to persist build/run commands, ports, and container conventions.\n\nWhen to use:\n- Any Dockerfile/compose changes or runtime env questions.\n\nOutputs:\n- Full updated Dockerfile/compose + rationale + how to run/verify.", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/docker.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "cicd": { > "mode": "subagent", > "model": "openrouter/mimo-flash", > "description": "CI/CD Pipeline Engineer.\n\nWhat it does:\n- Designs fast, deterministic, secure pipelines (GitHub Actions by default).\n- Minimizes privileges, maximizes caching.\n- Uses memory to persist canonical CI commands and pipeline conventions.\n\nWhen to use:\n- Any workflow/pipeline changes.\n\nOutputs:\n- Ready-to-apply YAML workflows (no TODO placeholders; unknowns marked as UNKNOWN with a discovery plan).", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/cicd.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "reviewer": { > "mode": "subagent", > "model": "openrouter/mimo-flash", > "description": "Adversarial Code Reviewer (read-only).\n\nWhat it does:\n- Reviews diffs for correctness, performance, readability.\n- Also checks: migrations/DB safety, backward compatibility, secrets/logging/PII.\n- Uses memory to keep project-specific review checklist (if stable).\n\nWhen to use:\n- Always after implementation before shipping.\n\nOutputs:\n- Actionable findings with file:line, issue, fix, why.", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/reviewer.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "security": { > "mode": "subagent", > "model": "openrouter/deepseek-r1t2", > "description": "Security / Threat Modeling (white hat).\n\nWhat it does:\n- Threat models and audits for common vuln classes (SQLi/XSS/SSRF/IDOR/supply-chain).\n- Produces exploit scenario + exact remediation.\n- Uses memory to persist known attack surfaces and security decisions.\n\nWhen to use:\n- Auth, secrets, input handling, network boundaries, permission systems.\n\nOutputs:\n- Findings categorized CRITICAL/HIGH/MEDIUM + verification checklist.", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/security.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "ideator": { > "mode": "subagent", > "model": "openrouter/r1t-chimera", > "description": "Product / Solution Ideation (Tree of Thoughts).\n\nWhat it does:\n- Generates MVP/Strategic/Moonshot concept branches.\n- Helps when requirements unclear or you want options.\n- Can store stable product constraints in memory if useful.\n\nWhen to use:\n- Early discovery, brainstorming, when user wants multiple approaches.\n\nOutputs:\n- 3 concepts with user story, feasibility, wow-factor.", > "temperature": 0.8, > "prompt": "{file:./instructions/inject/ideator.md}", > "tools": { > "read": true, > "skill": true, > "mcp": true > } > }, > > "docs": { > "mode": "subagent", > "model": "openrouter/mimo-flash", > "description": "Documentation / Developer Advocate.\n\nWhat it does:\n- Updates README and API docs so docs stay source-of-truth.\n- Ensures code blocks are copy/paste runnable.\n- Uses memory to persist canonical 'how to run/test' snippets.\n\nWhen to use:\n- Always when APIs change, new commands appear, or setup changes.\n\nOutputs:\n- List of files to update + full Markdown content.", > "temperature": 0.2, > "prompt": "{file:./instructions/inject/docs.md}", > "tools": { > "read": true, > "write": true, > "edit": true, > "skill": true, > "mcp": true > } > }, > > "version-scout": { > "mode": "subagent", > "model": "openrouter/mimo-flash", > "description": "Dependency / Version Researcher (latest stable).\n\nWhat it does:\n- Uses webfetch to find latest stable releases for libraries/tools/images.\n- Never guesses versions. If unknown → UNKNOWN.\n- Also checks if project is on an older version and calls it out.\n- Uses memory to store 'last verified version + date' (but re-verifies if 'latest' is requested).\n\nWhen to use:\n- Poetry deps, tool upgrades, when you suspect repo is behind.\n\nOutputs:\n- Markdown table: component, latest stable version, release date, source link, notes.", > "temperature": 0.1, > "prompt": "{file:./instructions/inject/version-scout.md}", > "tools": { > "read": true, > "webfetch": true, > "skill": true, > "mcp": true > } > } > }, > > "permission": { > "read": "allow", > "edit": "allow", > "write": "allow", > "webfetch": "allow", > > "skill": { "*": "allow" }, > > // 3.3 — фиксируем явно, чтобы поведение было читаемо в конфиге. > "doom_loop": "ask", > "external_directory": "ask", > > // 3.3 — делаем ask на явно опасные команды, остальное оставляем allow (как у тебя было). > "bash": { > "git commit*": "allow", > "git push": "ask", > > "rm *": "ask", > "sudo *": "ask", > "git reset *": "ask", > "git clean *": "ask", > "docker system prune*": "ask", > > "*": "allow" > }, > > "mcp": { > "context7": { > "type": "local", > "command": ["npx", "-y", "@upstash/context7-mcp@latest"], > "enabled": true, > "timeout": 60000 > }, > "sequential-thinking": { > "type": "local", > "command": ["npx", "-y", "@modelcontextprotocol/server-sequential-thinking"], > "enabled": true, > "timeout": 60000 > }, > "memory": { > "type": "local", > "command": ["npx", "-y", "@modelcontextprotocol/server-memory"], > "enabled": true, > "timeout": 60000 > } > } > } > } > ``` +1
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#2678
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?