test(config,tui): stop two tests reading the developer's environment and checkout path

`go test ./...` is red on a developer machine and green in CI, for two tests
that take an input from the machine rather than from the test.

TestResolveReportsExplicitMaxTurns left ResolveOptions.Env nil on four of its
five cases, and a nil Env means "read the real process environment" (envValue).
Any one of the ten variables the resolver reads is enough to break it. With
ANTHROPIC_BASE_URL exported, applyProviderEnv synthesizes a second, keyless
"anthropic" profile from a base URL on its own, the single-provider shortcut in
normalizeProviders no longer applies because there are now two, and Resolve
returns `no active provider configured: active provider "" not found` before the
test reaches its assertion. Measured one variable at a time against a clean
environment, all ten fail the same test: ANTHROPIC_BASE_URL, ANTHROPIC_API_KEY,
ANTHROPIC_MODEL, OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MODEL, GEMINI_API_KEY,
GOOGLE_API_KEY, ZERO_PROVIDER and ZERO_MAX_TURNS.

The cases now state an empty environment instead of no environment, which is
what they meant, and the test exports three hostile values itself so it proves
that rather than relying on the machine it runs on being clean. Reverting either
changed case fails the test on a clean environment too, so this cannot regress
silently on CI.

TestAltScreenTranscriptScrollKeepsFooterFixed asserted the branch survives in
the title bar at width 90 while leaving Options.Cwd empty, so newModel took the
real working directory. The outcome is not a monotonic function of that path's
length, which is why this took a measurement rather than a guess. The title bar
picks the widest candidate that fits: workspace carries branch and path, cwdOnly
carries the path alone, compactLeft carries the branch alone. At width 90 with
this branch, a cwd of up to 52 bytes keeps both, 53 to 74 falls to path-only and
drops the branch, and 75 or more falls again to branch-only and restores it. The
test failed from a checkout whose package path is 74 bytes and passed from one
at 80; CI sits at 40. Cwd is now pinned, which also stops newModel reading the
real repository for a branch the test overwrites anyway.

No production code changes. The resolver's nil-Env contract and the header's
candidate ladder are both deliberate; only the tests were leaning on them.

internal/config/resolver_test.go

@@ -2254,6 +2254,19 @@ func TestResolveRejectsInvalidCrossSessionInbound(t *testing.T) {
 }
 
 func TestResolveReportsExplicitMaxTurns(t *testing.T) {
+	// A nil ResolveOptions.Env means "read the real process environment" (see
+	// envValue), so a case that leaves it nil inherits whatever the developer
+	// happens to export. ANTHROPIC_BASE_URL alone is enough: applyProviderEnv
+	// synthesizes a second, keyless "anthropic" profile from a base URL on its
+	// own, the single-provider shortcut in normalizeProviders stops applying,
+	// and Resolve fails with `active provider "" not found` before this test
+	// reaches its assertion. Every variable the resolver reads does it, so the
+	// cases below state an EMPTY environment rather than no environment, and
+	// these three are set to a hostile value so the test proves it.
+	t.Setenv("ANTHROPIC_BASE_URL", "https://proxy.example/v1")
+	t.Setenv("OPENAI_API_KEY", "from-the-developer-shell")
+	t.Setenv(MaxTurnsEnv, "7")
+
 	const providerJSON = `{
 		"activeProvider": "p",
 		"providers": [{"name": "p", "provider_kind": "openai-compatible", "base_url": "https://x.example/v1", "apiKey": "k", "model": "m"}]
@@ -2266,10 +2279,15 @@ func TestResolveReportsExplicitMaxTurns(t *testing.T) {
 		set  bool
 	}{
 		{"built-in default only", ResolveOptions{
+			Env:       map[string]string{},
 			Overrides: Overrides{Provider: ProviderProfile{Name: "p", ProviderKind: "openai-compatible", BaseURL: "https://x.example/v1", APIKey: "k", Model: "m"}},
 		}, false},
-		{"user config", ResolveOptions{UserConfigPath: writeConfig(t, providerWithTurns)}, true},
+		{"user config", ResolveOptions{
+			Env:            map[string]string{},
+			UserConfigPath: writeConfig(t, providerWithTurns),
+		}, true},
 		{"project config", ResolveOptions{
+			Env:               map[string]string{},
 			UserConfigPath:    writeConfig(t, providerJSON),
 			ProjectConfigPath: writeConfig(t, `{"maxTurns": 40}`),
 		}, true},
@@ -2278,6 +2296,7 @@ func TestResolveReportsExplicitMaxTurns(t *testing.T) {
 			Overrides: Overrides{Provider: ProviderProfile{Name: "p", ProviderKind: "openai-compatible", BaseURL: "https://x.example/v1", APIKey: "k", Model: "m"}},
 		}, true},
 		{"cli override", ResolveOptions{
+			Env: map[string]string{},
 			Overrides: Overrides{
 				MaxTurns: 42,
 				Provider: ProviderProfile{Name: "p", ProviderKind: "openai-compatible", BaseURL: "https://x.example/v1", APIKey: "k", Model: "m"},

internal/providers/anthropic/provider.go

@@ -90,6 +90,7 @@ type Options struct {
 	// When unset, Zero uses providerio.ResolveStreamIdleTimeout — the
 	// ZERO_STREAM_IDLE_TIMEOUT override or providerio.DefaultStreamIdleTimeout.
 	StreamIdleTimeout time.Duration
+	SessionHeader     string
 }
 
 // Provider streams completions from Anthropic's Messages API.
@@ -108,6 +109,7 @@ type Provider struct {
 	httpClient        *http.Client
 	userAgent         string
 	streamIdleTimeout time.Duration
+	sessionHeader     providerio.SessionHeader
 }
 
 // New creates an Anthropic provider.
@@ -143,6 +145,7 @@ func New(options Options) (*Provider, error) {
 		httpClient:        providerio.HTTPClient(options.HTTPClient),
 		userAgent:         options.UserAgent,
 		streamIdleTimeout: providerio.ResolveStreamIdleTimeout(options.StreamIdleTimeout),
+		sessionHeader:     providerio.NewSessionHeader(options.SessionHeader),
 	}, nil
 }
 
@@ -163,12 +166,12 @@ func (provider *Provider) StreamCompletion(
 	events := make(chan zeroruntime.StreamEvent, 16)
 	go func() {
 		defer close(events)
-		provider.stream(ctx, body, events)
+		provider.stream(ctx, body, request.PromptCacheKey, events)
 	}()
 	return events, nil
 }
 
-func (provider *Provider) stream(ctx context.Context, body []byte, events chan<- zeroruntime.StreamEvent) {
+func (provider *Provider) stream(ctx context.Context, body []byte, sessionID string, events chan<- zeroruntime.StreamEvent) {
 	// streamCtx lets the idle watchdog abort an in-flight body read by cancelling
 	// the request, which unblocks the SSE reader goroutine.
 	streamCtx, cancelStream := context.WithCancel(ctx)
@@ -185,6 +188,7 @@ func (provider *Provider) stream(ctx context.Context, body []byte, events chan<-
 		},
 		provider.oauthResolver,
 		func(request *http.Request) {
+			provider.sessionHeader.Apply(request, sessionID)
 			request.Header.Set("Content-Type", "application/json")
 			request.Header.Set("anthropic-version", provider.version)
 			if provider.beta != "" {

internal/providers/factory.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"net/http"
+	"net/url"
 	"strings"
 
 	"github.com/Gitlawb/zero/internal/config"
@@ -53,6 +54,11 @@ func New(profile config.ProviderProfile, options Options) (zeroruntime.Provider,
 		return newCodexProvider(profile, resolved, options)
 	}
 
+	sessionHeader := sessionHeaderForProfile(profile, resolved)
+	if sessionHeader != "" && strings.TrimSpace(options.UserAgent) == "" {
+		options.UserAgent = "zero"
+	}
+
 	switch resolved.providerKind {
 	case config.ProviderKindOpenAI, config.ProviderKindOpenAICompatible:
 		// prompt_cache_key is an OpenAI-only chat-completions field. Strict
@@ -74,6 +80,7 @@ func New(profile config.ProviderProfile, options Options) (zeroruntime.Provider,
 			UserAgent:             options.UserAgent,
 			ParseThinkTags:        parseThinkTagsForProfile(profile, resolved),
 			DisablePromptCacheKey: resolved.providerKind == config.ProviderKindOpenAICompatible,
+			SessionHeader:         sessionHeader,
 		})
 	case config.ProviderKindAnthropic, config.ProviderKindAnthropicCompat:
 		return anthropic.New(anthropic.Options{
@@ -88,6 +95,7 @@ func New(profile config.ProviderProfile, options Options) (zeroruntime.Provider,
 			MaxTokens:       resolved.maxOutputTokens,
 			HTTPClient:      options.HTTPClient,
 			UserAgent:       options.UserAgent,
+			SessionHeader:   sessionHeader,
 		})
 	case config.ProviderKindGoogle:
 		return gemini.New(gemini.Options{
@@ -345,6 +353,18 @@ func isCodexCatalog(profile config.ProviderProfile, _ resolvedProfile) bool {
 	return providercatalog.NormalizeID(profile.CatalogID) == "chatgpt"
 }
 
+func sessionHeaderForProfile(profile config.ProviderProfile, resolved resolvedProfile) string {
+	switch providercatalog.NormalizeID(profile.CatalogID) {
+	case "opencode", "opencode-go", "opencode-go-anthropic-compatible":
+		return "x-opencode-session"
+	}
+	endpoint, err := url.Parse(resolved.baseURL)
+	if err == nil && strings.EqualFold(endpoint.Hostname(), "opencode.ai") && (endpoint.Path == "/zen" || strings.HasPrefix(endpoint.Path, "/zen/")) {
+		return "x-opencode-session"
+	}
+	return ""
+}
+
 // newCodexProvider builds a Codex-flavored openai provider for the chatgpt
 // catalog. The Codex headers (`originator`, `chatgpt-account-id`) are
 // injected by the CodexProvider wrapper. The `chatgpt-account-id` is read from

internal/providers/factory_test.go

@@ -3,13 +3,17 @@ package providers
 import (
 	"context"
 	"encoding/json"
+	"fmt"
 	"io"
 	"net/http"
 	"strings"
+	"sync"
+	"sync/atomic"
 	"testing"
 
 	"github.com/Gitlawb/zero/internal/config"
 	"github.com/Gitlawb/zero/internal/oauth"
+	"github.com/Gitlawb/zero/internal/providers/providerio"
 	"github.com/Gitlawb/zero/internal/zeroruntime"
 )
 
@@ -794,3 +798,566 @@ func TestResolveRuntimeMetadataRejectsModelOutsideProviderAllowlist(t *testing.T
 		t.Fatalf("error = %q, want it to name the rejected model", err.Error())
 	}
 }
+
+func TestOpenCodeSessionHeader(t *testing.T) {
+	const sessionHeader = "x-opencode-session"
+	tests := []struct {
+		name       string
+		profile    config.ProviderProfile
+		response   string
+		wantHeader bool
+		wantAuth   map[string]string
+	}{
+		{
+			name: "catalog opencode-go default URL",
+			profile: config.ProviderProfile{
+				Name:         "opencode-go",
+				CatalogID:    "opencode-go",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				APIKey:       "sk-oc-go",
+				Model:        "deepseek-v4-pro",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: true,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-oc-go"},
+		},
+		{
+			name: "catalog opencode-go-anthropic-compatible",
+			profile: config.ProviderProfile{
+				Name:         "opencode-go-anthropic",
+				CatalogID:    "opencode-go-anthropic-compatible",
+				ProviderKind: config.ProviderKindAnthropicCompat,
+				APIKey:       "sk-oc-ant",
+				Model:        "minimax-m3",
+			},
+			response:   "data: {\"type\":\"message_stop\"}\n\n",
+			wantHeader: true,
+			wantAuth:   map[string]string{"x-api-key": "sk-oc-ant"},
+		},
+		{
+			name: "catalog opencode Zen",
+			profile: config.ProviderProfile{
+				Name:         "opencode-zen",
+				CatalogID:    "opencode",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				APIKey:       "sk-zen",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: true,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-zen"},
+		},
+		{
+			name: "renamed catalog profile with custom proxy URL",
+			profile: config.ProviderProfile{
+				Name:         "my-opencode",
+				CatalogID:    "opencode-go",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				BaseURL:      "https://proxy.example/v1",
+				APIKey:       "sk-proxy",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: true,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-proxy"},
+		},
+		{
+			name: "catalog-less OpenAI-compatible OpenCode Go URL",
+			profile: config.ProviderProfile{
+				Name:         "manual-go",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				BaseURL:      "https://opencode.ai/zen/go/v1",
+				APIKey:       "sk-manual",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: true,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-manual"},
+		},
+		{
+			name: "catalog-less Anthropic-compatible OpenCode Go URL",
+			profile: config.ProviderProfile{
+				Name:         "manual-go-anthropic",
+				ProviderKind: config.ProviderKindAnthropicCompat,
+				BaseURL:      "https://opencode.ai/zen/go",
+				APIKey:       "sk-manual-ant",
+				Model:        "test-model",
+			},
+			response:   "data: {\"type\":\"message_stop\"}\n\n",
+			wantHeader: true,
+			wantAuth:   map[string]string{"x-api-key": "sk-manual-ant"},
+		},
+		{
+			name: "official OpenAI",
+			profile: config.ProviderProfile{
+				Name:         "openai",
+				ProviderKind: config.ProviderKindOpenAI,
+				APIKey:       "sk-openai",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: false,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-openai"},
+		},
+		{
+			name: "generic OpenAI-compatible",
+			profile: config.ProviderProfile{
+				Name:         "generic",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				BaseURL:      "https://provider.example/v1",
+				APIKey:       "sk-generic",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: false,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-generic"},
+		},
+		{
+			name: "official Anthropic",
+			profile: config.ProviderProfile{
+				Name:         "anthropic",
+				ProviderKind: config.ProviderKindAnthropic,
+				APIKey:       "sk-ant",
+				Model:        "test-model",
+			},
+			response:   "data: {\"type\":\"message_stop\"}\n\n",
+			wantHeader: false,
+			wantAuth:   map[string]string{"x-api-key": "sk-ant"},
+		},
+		{
+			name: "lookalike subdomain is not OpenCode",
+			profile: config.ProviderProfile{
+				Name:         "deceptive-host",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				BaseURL:      "https://opencode.ai.example/zen/go/v1",
+				APIKey:       "sk-fake",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: false,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-fake"},
+		},
+		{
+			name: "lookalike path is not OpenCode",
+			profile: config.ProviderProfile{
+				Name:         "deceptive-path",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				BaseURL:      "https://example.com/opencode.ai/zen/go",
+				APIKey:       "sk-fake",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: false,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-fake"},
+		},
+		{
+			name: "zenith path is not Zen",
+			profile: config.ProviderProfile{
+				Name:         "zenith",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				BaseURL:      "https://opencode.ai/zenith",
+				APIKey:       "sk-zenith",
+				Model:        "test-model",
+			},
+			response:   "data: [DONE]\n\n",
+			wantHeader: false,
+			wantAuth:   map[string]string{"Authorization": "Bearer sk-zenith"},
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			transport := &captureTransport{responseBody: test.response}
+			provider, err := New(test.profile, Options{
+				HTTPClient: &http.Client{Transport: transport},
+				UserAgent:  "zero-test",
+			})
+			if err != nil {
+				t.Fatalf("New() error = %v", err)
+			}
+			stream, err := provider.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+				Messages:       []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+				PromptCacheKey: "conversation-a",
+			})
+			if err != nil {
+				t.Fatalf("StreamCompletion() error = %v", err)
+			}
+			drainStream(t, stream)
+			if transport.request == nil {
+				t.Fatal("HTTP client was not used")
+			}
+			got := transport.request.Header.Get(sessionHeader)
+			if test.wantHeader && got != "conversation-a" {
+				t.Fatalf("%s = %q, want %q", sessionHeader, got, "conversation-a")
+			}
+			if !test.wantHeader && got != "" {
+				t.Fatalf("%s = %q, want no session header", sessionHeader, got)
+			}
+			for header, want := range test.wantAuth {
+				if got := transport.request.Header.Get(header); got != want {
+					t.Fatalf("%s = %q, want %q", header, got, want)
+				}
+			}
+			if got := transport.request.Header.Get("User-Agent"); got != "zero-test" {
+				t.Fatalf("User-Agent = %q, want zero-test", got)
+			}
+			if test.profile.ProviderKind == config.ProviderKindOpenAICompatible {
+				var body map[string]any
+				if err := json.NewDecoder(transport.body()).Decode(&body); err != nil {
+					t.Fatalf("decode request body: %v", err)
+				}
+				if _, ok := body["prompt_cache_key"]; ok {
+					t.Fatalf("openai-compatible body carried prompt_cache_key: %#v", body["prompt_cache_key"])
+				}
+			}
+		})
+	}
+}
+
+func TestOpenCodeSessionHeaderFollowsConversationKey(t *testing.T) {
+	transport := &recordingTransport{status: http.StatusOK, body: "data: [DONE]\n\n"}
+	profile := config.ProviderProfile{
+		Name:         "opencode-go",
+		CatalogID:    "opencode-go",
+		ProviderKind: config.ProviderKindOpenAICompatible,
+		APIKey:       "sk-oc",
+		Model:        "test-model",
+	}
+	options := Options{HTTPClient: &http.Client{Transport: transport}, UserAgent: "zero-test"}
+	provider, err := New(profile, options)
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	for _, key := range []string{"conversation-a", "conversation-a", "conversation-b"} {
+		stream, err := provider.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+			Messages:       []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+			PromptCacheKey: key,
+		})
+		if err != nil {
+			t.Fatalf("StreamCompletion() error = %v", err)
+		}
+		drainStream(t, stream)
+	}
+	got := transport.sessionHeaders()
+	want := []string{"conversation-a", "conversation-a", "conversation-b"}
+	if len(got) != len(want) {
+		t.Fatalf("captured %d requests, want %d", len(got), len(want))
+	}
+	for i := range want {
+		if got[i] != want[i] {
+			t.Fatalf("request %d x-opencode-session = %q, want %q", i, got[i], want[i])
+		}
+	}
+
+	reconstructed, err := New(profile, options)
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	stream, err := reconstructed.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+		Messages:       []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+		PromptCacheKey: "conversation-a",
+	})
+	if err != nil {
+		t.Fatalf("StreamCompletion() error = %v", err)
+	}
+	drainStream(t, stream)
+	got = transport.sessionHeaders()
+	if last := got[len(got)-1]; last != "conversation-a" {
+		t.Fatalf("reconstructed provider sent %s = %q, want conversation-a", "x-opencode-session", last)
+	}
+}
+
+func TestOpenCodeSessionHeaderCustomOverride(t *testing.T) {
+	custom := map[string]string{"X-Opencode-Session": "explicit-session"}
+	profile := config.ProviderProfile{
+		Name:          "opencode-go",
+		CatalogID:     "opencode-go",
+		ProviderKind:  config.ProviderKindOpenAICompatible,
+		APIKey:        "sk-oc",
+		Model:         "test-model",
+		CustomHeaders: custom,
+	}
+	transport := &captureTransport{responseBody: "data: [DONE]\n\n"}
+	provider, err := New(profile, Options{HTTPClient: &http.Client{Transport: transport}, UserAgent: "zero-test"})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	stream, err := provider.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+		Messages:       []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+		PromptCacheKey: "conversation-a",
+	})
+	if err != nil {
+		t.Fatalf("StreamCompletion() error = %v", err)
+	}
+	drainStream(t, stream)
+	if got := transport.request.Header.Get("x-opencode-session"); got != "explicit-session" {
+		t.Fatalf("x-opencode-session = %q, want explicit custom value", got)
+	}
+	if len(custom) != 1 || custom["X-Opencode-Session"] != "explicit-session" {
+		t.Fatalf("profile CustomHeaders mutated: %#v", custom)
+	}
+}
+
+func TestOpenCodeSessionHeaderFallback(t *testing.T) {
+	transport := &recordingTransport{status: http.StatusOK, body: "data: [DONE]\n\n"}
+	profile := config.ProviderProfile{
+		Name:         "opencode-go",
+		CatalogID:    "opencode-go",
+		ProviderKind: config.ProviderKindOpenAICompatible,
+		APIKey:       "sk-oc",
+		Model:        "test-model",
+	}
+	options := Options{HTTPClient: &http.Client{Transport: transport}}
+	first, err := New(profile, options)
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	second, err := New(profile, options)
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	request := zeroruntime.CompletionRequest{
+		Messages: []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+	}
+	for _, provider := range []zeroruntime.Provider{first, first, second} {
+		stream, err := provider.StreamCompletion(context.Background(), request)
+		if err != nil {
+			t.Fatalf("StreamCompletion() error = %v", err)
+		}
+		drainStream(t, stream)
+	}
+	headers := transport.sessionHeaders()
+	if len(headers) != 3 {
+		t.Fatalf("captured %d requests, want 3", len(headers))
+	}
+	if headers[0] == "" || headers[1] != headers[0] {
+		t.Fatalf("fallback headers = %q, %q — want one stable nonempty value per provider", headers[0], headers[1])
+	}
+	if headers[2] == headers[0] {
+		t.Fatalf("distinct provider instances shared fallback %q", headers[0])
+	}
+}
+
+func TestOpenCodeSessionHeaderDefaultsUserAgent(t *testing.T) {
+	transport := &captureTransport{responseBody: "data: [DONE]\n\n"}
+	provider, err := New(config.ProviderProfile{
+		Name:         "opencode-go",
+		CatalogID:    "opencode-go",
+		ProviderKind: config.ProviderKindOpenAICompatible,
+		APIKey:       "sk-oc",
+		Model:        "test-model",
+	}, Options{HTTPClient: &http.Client{Transport: transport}})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	stream, err := provider.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+		Messages: []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+	})
+	if err != nil {
+		t.Fatalf("StreamCompletion() error = %v", err)
+	}
+	drainStream(t, stream)
+	if got := transport.request.Header.Get("User-Agent"); got != "zero" {
+		t.Fatalf("User-Agent = %q, want zero", got)
+	}
+}
+
+func TestOpenCodeSessionHeaderConcurrent(t *testing.T) {
+	transport := &recordingTransport{status: http.StatusOK, body: "data: [DONE]\n\n"}
+	provider, err := New(config.ProviderProfile{
+		Name:         "opencode-go",
+		CatalogID:    "opencode-go",
+		ProviderKind: config.ProviderKindOpenAICompatible,
+		APIKey:       "sk-oc",
+		Model:        "test-model",
+	}, Options{HTTPClient: &http.Client{Transport: transport}})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+
+	var wg sync.WaitGroup
+	var errsMu sync.Mutex
+	var errs []string
+	for _, session := range []string{"session-a", "session-b"} {
+		wg.Add(1)
+		go func(session string) {
+			defer wg.Done()
+			for i := 0; i < 5; i++ {
+				stream, err := provider.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+					Messages:       []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: fmt.Sprintf("req-%s-%d", session, i)}},
+					PromptCacheKey: session,
+				})
+				if err != nil {
+					errsMu.Lock()
+					errs = append(errs, err.Error())
+					errsMu.Unlock()
+					return
+				}
+				for event := range stream {
+					if event.Type == zeroruntime.StreamEventError {
+						errsMu.Lock()
+						errs = append(errs, event.Error)
+						errsMu.Unlock()
+					}
+				}
+			}
+		}(session)
+	}
+	wg.Wait()
+	for _, errText := range errs {
+		t.Errorf("concurrent stream error: %s", errText)
+	}
+
+	requests := transport.recorded()
+	if len(requests) != 10 {
+		t.Fatalf("captured %d requests, want 10", len(requests))
+	}
+	for _, request := range requests {
+		var payload struct {
+			Messages []struct {
+				Content string `json:"content"`
+			} `json:"messages"`
+		}
+		if err := json.Unmarshal([]byte(request.body), &payload); err != nil {
+			t.Fatalf("decode request body: %v", err)
+		}
+		if len(payload.Messages) == 0 {
+			t.Fatal("request carried no user message")
+		}
+		content := payload.Messages[0].Content
+		want := "session-a"
+		if strings.Contains(content, "session-b") {
+			want = "session-b"
+		}
+		if got := request.header.Get("x-opencode-session"); got != want {
+			t.Fatalf("request for %q carried x-opencode-session %q, want %q", content, got, want)
+		}
+	}
+}
+
+func TestOpenCodeSessionHeaderAuthRetry(t *testing.T) {
+	for _, test := range []struct {
+		name     string
+		profile  config.ProviderProfile
+		response string
+	}{
+		{
+			name: "openai-compatible",
+			profile: config.ProviderProfile{
+				Name:         "opencode-go",
+				CatalogID:    "opencode-go",
+				ProviderKind: config.ProviderKindOpenAICompatible,
+				APIKey:       "sk-oc",
+				Model:        "test-model",
+			},
+			response: "data: [DONE]\n\n",
+		},
+		{
+			name: "anthropic-compatible",
+			profile: config.ProviderProfile{
+				Name:         "opencode-go-anthropic",
+				CatalogID:    "opencode-go-anthropic-compatible",
+				ProviderKind: config.ProviderKindAnthropicCompat,
+				APIKey:       "sk-oc-ant",
+				Model:        "minimax-m3",
+			},
+			response: "data: {\"type\":\"message_stop\"}\n\n",
+		},
+	} {
+		t.Run(test.name, func(t *testing.T) {
+			transport := &recordingTransport{status: http.StatusOK, body: test.response, unauthorized: 1}
+			var calls atomic.Int32
+			resolver := providerio.TokenResolver(func(context.Context, bool) (string, string, bool, error) {
+				n := calls.Add(1)
+				return "Authorization", fmt.Sprintf("Bearer tok-%d", n), true, nil
+			})
+			provider, err := New(test.profile, Options{
+				HTTPClient:    &http.Client{Transport: transport},
+				OAuthResolver: resolver,
+			})
+			if err != nil {
+				t.Fatalf("New() error = %v", err)
+			}
+			stream, err := provider.StreamCompletion(context.Background(), zeroruntime.CompletionRequest{
+				Messages:       []zeroruntime.Message{{Role: zeroruntime.MessageRoleUser, Content: "hello"}},
+				PromptCacheKey: "conversation-a",
+			})
+			if err != nil {
+				t.Fatalf("StreamCompletion() error = %v", err)
+			}
+			drainStream(t, stream)
+			requests := transport.recorded()
+			if len(requests) != 2 {
+				t.Fatalf("captured %d requests, want 2 (401 + retry)", len(requests))
+			}
+			for i, request := range requests {
+				if got := request.header.Get("x-opencode-session"); got != "conversation-a" {
+					t.Fatalf("attempt %d x-opencode-session = %q, want conversation-a", i, got)
+				}
+			}
+			if first, second := requests[0].header.Get("Authorization"), requests[1].header.Get("Authorization"); first == second {
+				t.Fatalf("Authorization not refreshed across retry: %q", first)
+			}
+		})
+	}
+}
+
+func drainStream(t *testing.T, stream <-chan zeroruntime.StreamEvent) {
+	t.Helper()
+	for event := range stream {
+		if event.Type == zeroruntime.StreamEventError {
+			t.Fatalf("stream error: %s", event.Error)
+		}
+	}
+}
+
+type recordedRequest struct {
+	header http.Header
+	body   string
+}
+
+type recordingTransport struct {
+	mu           sync.Mutex
+	requests     []recordedRequest
+	status       int
+	body         string
+	unauthorized int
+}
+
+func (transport *recordingTransport) RoundTrip(request *http.Request) (*http.Response, error) {
+	var body string
+	if request.Body != nil {
+		data, _ := io.ReadAll(request.Body)
+		body = string(data)
+	}
+	transport.mu.Lock()
+	transport.requests = append(transport.requests, recordedRequest{header: request.Header.Clone(), body: body})
+	attempt := len(transport.requests)
+	transport.mu.Unlock()
+
+	status := transport.status
+	if attempt <= transport.unauthorized {
+		status = http.StatusUnauthorized
+	}
+	return &http.Response{
+		StatusCode: status,
+		Status:     fmt.Sprintf("%d", status),
+		Header:     make(http.Header),
+		Body:       io.NopCloser(strings.NewReader(transport.body)),
+		Request:    request,
+	}, nil
+}
+
+func (transport *recordingTransport) recorded() []recordedRequest {
+	transport.mu.Lock()
+	defer transport.mu.Unlock()
+	return append([]recordedRequest(nil), transport.requests...)
+}
+
+func (transport *recordingTransport) sessionHeaders() []string {
+	requests := transport.recorded()
+	headers := make([]string, 0, len(requests))
+	for _, request := range requests {
+		headers = append(headers, request.header.Get("x-opencode-session"))
+	}
+	return headers
+}

internal/providers/openai/provider.go

@@ -65,6 +65,7 @@ type Options struct {
 	// enabled; the ZERO_DISABLE_PROMPT_CACHE_KEY env kill switch still applies
 	// on top for any endpoint.
 	DisablePromptCacheKey bool
+	SessionHeader         string
 }
 
 // Provider streams completions from an OpenAI-compatible chat completions API.
@@ -85,6 +86,7 @@ type Provider struct {
 	parseThinkTags        bool
 	setRequestExtra       func(*http.Request)
 	disablePromptCacheKey bool
+	sessionHeader         providerio.SessionHeader
 }
 
 // New creates an OpenAI-compatible provider.
@@ -140,6 +142,7 @@ func New(options Options) (*Provider, error) {
 		parseThinkTags:        options.ParseThinkTags,
 		setRequestExtra:       options.SetRequestExtra,
 		disablePromptCacheKey: options.DisablePromptCacheKey,
+		sessionHeader:         providerio.NewSessionHeader(options.SessionHeader),
 	}, nil
 }
 
@@ -156,13 +159,13 @@ func (provider *Provider) StreamCompletion(
 	events := make(chan zeroruntime.StreamEvent, 16)
 	go func() {
 		defer close(events)
-		provider.stream(ctx, body, events)
+		provider.stream(ctx, body, request.PromptCacheKey, events)
 	}()
 
 	return events, nil
 }
 
-func (provider *Provider) stream(ctx context.Context, body []byte, events chan<- zeroruntime.StreamEvent) {
+func (provider *Provider) stream(ctx context.Context, body []byte, sessionID string, events chan<- zeroruntime.StreamEvent) {
 	endpoint := provider.endpoint
 
 	// streamCtx lets the idle watchdog abort an in-flight body read by cancelling
@@ -186,6 +189,7 @@ func (provider *Provider) stream(ctx context.Context, body []byte, events chan<-
 		},
 		provider.oauthResolver,
 		func(request *http.Request) {
+			provider.sessionHeader.Apply(request, sessionID)
 			request.Header.Set("Content-Type", "application/json")
 			if provider.userAgent != "" {
 				request.Header.Set("User-Agent", provider.userAgent)

internal/providers/providerio/headers.go

@@ -1,6 +1,7 @@
 package providerio
 
 import (
+	"crypto/rand"
 	"net/http"
 	"strings"
 )
@@ -52,6 +53,28 @@ func ApplyAuthHeaders(request *http.Request, options AuthHeaders) {
 	request.Header.Set(header, value)
 }
 
+type SessionHeader struct {
+	name     string
+	fallback string
+}
+
+func NewSessionHeader(name string) SessionHeader {
+	if name == "" {
+		return SessionHeader{}
+	}
+	return SessionHeader{name: name, fallback: rand.Text()}
+}
+
+func (header SessionHeader) Apply(request *http.Request, sessionID string) {
+	if header.name == "" {
+		return
+	}
+	if sessionID = strings.TrimSpace(sessionID); sessionID == "" {
+		sessionID = header.fallback
+	}
+	request.Header.Set(header.name, sessionID)
+}
+
 func CopyHeaders(headers map[string]string) map[string]string {
 	if headers == nil {
 		return nil

internal/tui/scroll_test.go

@@ -2,6 +2,7 @@ package tui
 
 import (
 	"context"
+	"path/filepath"
 	"strings"
 	"testing"
 
@@ -112,7 +113,22 @@ func TestMouseWheelOnClippedFooterStatusDoesNotMoveComposerCursor(t *testing.T)
 }
 
 func TestAltScreenTranscriptScrollKeepsFooterFixed(t *testing.T) {
-	m := newModel(context.Background(), Options{AltScreen: true, ProviderName: "openai", ModelName: "gpt-4.1"})
+	// Cwd is pinned because an empty Options.Cwd makes newModel take the real
+	// working directory, which at test time is this package inside whoever's
+	// checkout, and the branch assertion below is not a monotonic function of
+	// that path's length. The title bar picks the widest candidate that fits:
+	// workspace carries branch AND path, cwdOnly carries only the path, and
+	// compactLeft carries only the branch. At width 90 with this branch, a cwd
+	// of up to 52 bytes keeps branch and path, 53 to 74 falls to path-only and
+	// loses the branch, and 75 or more falls again to branch-only and gets it
+	// back. So this failed from a checkout 74 bytes deep and passed from one 80
+	// bytes deep, while CI sits at 40 and never sees it.
+	m := newModel(context.Background(), Options{
+		AltScreen:    true,
+		ProviderName: "openai",
+		ModelName:    "gpt-4.1",
+		Cwd:          filepath.Join(string(filepath.Separator), "workspace", "zero"),
+	})
 	m.width = 90
 	m.height = 10
 	m.gitBranch = "feat/pinned-header"