Two prefixes: provider settings use short names (anthropic.apiKey), feature settings use codebuddy.*. Open in the editor with Cmd/Ctrl + , → filter by codebuddy or provider name.
Note for maintainers: this page is meant to be regenerated from package.json contributes.configuration on every release. Hand-maintained lists rot — every setting rename or default change requires touching this file. A build-time generator would be rot-proof; tracked as follow-up.
| Setting | Type | Default | Values |
|---|
generativeAi.option | enum | null | Active provider — Anthropic, OpenAI, Gemini, Groq, XGrok, Deepseek, Qwen, GLM, Local. null on first install → onboarding wizard fires |
| Setting | Type | Default | Description |
|---|
google.gemini.apiKeys | string | — | Gemini API key |
google.gemini.model | string | "gemini-2.5-pro" | Gemini model name |
| Setting | Type | Default | Description |
|---|
groq.llama3.apiKey | string | — | Groq API key |
groq.llama3.model | string | "llama-3.1-70b-versatile" | Groq model name |
| Setting | Type | Default | Description |
|---|
anthropic.apiKey | string | — | Anthropic API key |
anthropic.model | string | "claude-sonnet-4-5" | Anthropic model name |
| Setting | Type | Default | Description |
|---|
openai.apiKey | string | — | OpenAI API key |
openai.model | string | "gpt-4o" | OpenAI model name |
| Setting | Type | Default | Description |
|---|
deepseek.apiKey | string | — | Deepseek API key |
deepseek.model | string | "deepseek-chat" | Deepseek model (e.g., deepseek-chat V3, deepseek-reasoner R1) |
| Setting | Type | Default | Description |
|---|
qwen.apiKey | string | — | Qwen (DashScope) API key |
qwen.model | string | "qwen-max" | Qwen model (e.g., qwen-max, qwen-plus, qwen3-coder-plus) |
| Setting | Type | Default | Description |
|---|
glm.apiKey | string | — | GLM (Zhipu AI) API key |
glm.model | string | "glm-4" | GLM model (e.g., glm-4, glm-4-plus, glm-4v) |
| Setting | Type | Default | Description |
|---|
local.model | string | "qwen2.5-coder" | Local model name (e.g., qwen2.5-coder, llama3.2, gemma3) |
local.baseUrl | string | "http://localhost:11434/v1" | Base URL. Ollama: http://localhost:11434/v1, Docker Model Runner: http://localhost:12434/engines/llama.cpp/v1 |
local.apiKey | string | "not-needed" | API key for local LLM (if required) |
The web_search tool uses whichever key is set (Tavily, Brave, or Firecrawl). Gemini keys also power embeddings.
| Setting | Type | Default | Description |
|---|
tavily.apiKey | string | null | Tavily API key for web search |
brave.apiKey | string | null | Brave Search API key |
firecrawl.apiKey | string | null | Firecrawl API key |
| Setting | Type | Default | Description |
|---|
codebuddy.codeBuddyMode | enum | "Agent" | Active interaction mode: Agent, Plan, Ask |
codebuddy.enableStreaming | boolean | true | Enable streaming responses from AI models |
codebuddy.autoApprove | boolean | false | Automatically approve agent actions without confirmation |
codebuddy.allowFileEdits | boolean | true | Allow the agent to create, modify, and delete files |
codebuddy.allowTerminal | boolean | true | Allow the agent to execute terminal commands |
codebuddy.verboseLogging | boolean | false | Show detailed agent activity logs for debugging |
codebuddy.indexCodebase | boolean | false | Enable vector database indexing for semantic code search |
codebuddy.contextWindow | enum | "16k" | Max context window size: 4k, 8k, 16k, 32k, 128k |
codebuddy.includeHidden | boolean | false | Include hidden files (.-prefixed) in context gathering |
codebuddy.maxFileSize | string | "1" | Max file size in MB for context gathering |
codebuddy.requireDiffApproval | boolean | false | Require manual approval before file changes are applied |
codebuddy.language | enum | "en" | UI language: en, es, fr, de, zh-cn, ja, yo |
codebuddy.browserType | enum | "system" | How to open URLs: reader, simple, system |
codebuddy.nickname | string | "" | Your display name in chat conversations |
codebuddy.compactMode | boolean | false | Reduce spacing between messages for a denser view |
codebuddy.showTimestamps | boolean | false | Show a timestamp under each chat message |
codebuddy.sessionAwareness.enabled | boolean | true | Tell the agent which files it already read this session so it doesn’t re-read |
codebuddy.hooks.enabled | boolean | true | Run pre/post tool-execution hooks from .codebuddy/hooks.json |
codebuddy.debug.fileLogging | boolean | false | Write logs to .codebuddy/logs (auto-on when running from source) |
| Setting | Type | Default | Description |
|---|
font.family | enum | "JetBrains Mono" | Chat font family: Montserrat, SF Mono, Space Mono, Fira Code, Source Code Pro, JetBrains Mono, Roboto Mono, Ubuntu Mono, IBM Plex Mono, Inconsolata |
chatview.theme | enum | "Atom One Dark" | Syntax highlight theme: Atom One Dark, Code Pen, github dark, night owl, tokyo night, and others |
chatview.font.size | number | 16 | Chat font size in pixels |
| Setting | Type | Default | Range | Description |
|---|
codebuddy.agent.maxEventCount | number | 15000 | 500–50000 | Max stream events before the agent pauses |
codebuddy.agent.maxToolInvocations | number | 2000 | 50–10000 | Max total tool calls before the agent pauses |
codebuddy.agent.maxDurationMinutes | number | 60 | 1–240 | Max wall-clock minutes before the agent pauses |
codebuddy.agent.recursionLimit | number | 600 | 50–2000 | Max LangGraph recursion depth before force-stop |
codebuddy.agent.maxToolCallsPerType | number | 150 | 5–1000 | Per-tool cap before it’s treated as looping |
codebuddy.agent.fileEditLoopThreshold | number | 20 | 2–100 | Consecutive same-file edits before it’s treated as looping |
codebuddy.agent.maxConcurrentStreams | number | 3 | 1–10 | Max concurrent agent streams; additional requests are queued |
codebuddy.agent.contextBudgetWarning | boolean | true | — | Warn the agent when its context window passes ~70% so it summarizes |
| Setting | Type | Default | Range |
|---|
codebuddy.agent.limits.editFile | number | 75 | 1–400 |
codebuddy.agent.limits.writeFile | number | 75 | 1–400 |
codebuddy.agent.limits.deleteFile | number | 30 | 1–200 |
codebuddy.agent.limits.runCommand | number | 100 | 1–400 |
codebuddy.agent.limits.runTerminalCommand | number | 500 | 10–4000 |
codebuddy.agent.limits.webSearch | number | 60 | 1–400 |
| Setting | Type | Default | Range | Description |
|---|
codebuddy.agent.rateLimiter.maxTotalPerWindow | number | 600 | 60–3000 | Max tool calls across all tools/min |
codebuddy.agent.rateLimiter.maxPerToolPerWindow | number | 200 | 10–1000 | Max calls for any single tool/min |
| Setting | Type | Default | Description |
|---|
codebuddy.agent.costThreshold | number | 1000 | Max spend (USD) per conversation before prompting to continue |
codebuddy.agent.costThresholdWarningPercent | number | 80 | Percent of threshold at which a warning shows (1–99) |
codebuddy.agent.dailySpendCap | number/null | null | Max total daily spend (USD); null disables |
| Setting | Type | Default | Description |
|---|
codebuddy.agent.includeGitContext | boolean | true | Inject git branch/status/recent diffs into the system prompt at session start |
codebuddy.agent.groundingContext | boolean | false | First-turn: inject vector-retrieved code so the agent starts grounded. Needs a working embeddings provider; default OFF |
codebuddy.agent.changeAwareStart | boolean | false | First-turn: inject a compact list of files changed since the last session |
codebuddy.agent.anthropic.thinkingBudget | number | 0 | Anthropic extended-thinking budget in tokens (0 = off; enable ≥1024, rec. 2048–8192) |
| Setting | Type | Default | Description |
|---|
codebuddy.completion.enabled | boolean | true | Enable inline code completions (ghost text) |
codebuddy.completion.provider | enum | "Local" | Completion provider: Gemini, Groq, Anthropic, Deepseek, OpenAI, Qwen, GLM, Local |
codebuddy.completion.model | string | "qwen2.5-coder" | Model to use for completions |
codebuddy.completion.apiKey | string | "" | API key for completion provider (falls back to main key if empty) |
codebuddy.completion.debounceMs | number | 300 | Delay in ms before triggering a completion (min: 50) |
codebuddy.completion.maxTokens | number | 128 | Maximum tokens to generate per completion |
codebuddy.completion.triggerMode | enum | "automatic" | Trigger mode: automatic, manual |
codebuddy.completion.multiLine | boolean | true | Allow multi-line completions |
| Setting | Type | Default | Description |
|---|
codebuddy.rules.enabled | boolean | true | Enable project rules from .codebuddy/rules.md |
codebuddy.rules.maxTokens | number | 2000 | Max tokens for project rules (truncates if exceeded) |
codebuddy.rules.showIndicator | boolean | true | Show indicator when project rules are active |
rules.customRules | array | [] | Custom rules for code generation (objects with id, name, description, content, enabled) |
rules.customSystemPrompt | string | "" | Additional instructions appended to the base system prompt |
rules.subagents | object | {} | Per-subagent toggle configuration. Keys: subagent ID, values: { enabled: boolean } |
| Setting | Type | Default | Description |
|---|
codebuddy.review.inlineComments | boolean | true | Show code review comments inline in workspace files |
| Setting | Type | Default | Description |
|---|
codebuddy.testCommand | string | "" | Custom test command. If empty, auto-detects (Jest, Vitest, Mocha, Pytest, Go, Cargo) |
codebuddy.testTimeout | number | 120000 | Test execution timeout in ms (5000–600000) |
| Setting | Type | Default | Description |
|---|
codebuddy.vectorDb.enabled | boolean | true | Enable vector database for semantic code search |
codebuddy.vectorDb.embeddingModel | enum | "gemini" | Embedding provider: gemini, openai, local |
codebuddy.vectorDb.maxTokens | number | 6000 | Max tokens per chunk |
codebuddy.vectorDb.batchSize | number | 10 | Files per embedding batch (1–50) |
codebuddy.vectorDb.searchResultLimit | number | 8 | Max search results (1–20) |
codebuddy.vectorDb.enableBackgroundProcessing | boolean | true | Index changes in the background |
codebuddy.vectorDb.enableProgressNotifications | boolean | true | Show indexing progress |
codebuddy.vectorDb.progressLocation | enum | "notification" | Progress location: notification, statusBar |
codebuddy.vectorDb.debounceDelay | number | 1000 | Re-index debounce delay in ms |
codebuddy.vectorDb.performanceMode | enum | "balanced" | Performance mode: balanced, performance, memory |
codebuddy.vectorDb.fallbackToKeywordSearch | boolean | true | Fall back to keyword search when vectors fail |
codebuddy.vectorDb.cacheEnabled | boolean | true | Cache search results |
codebuddy.vectorDb.logLevel | enum | "info" | Log level: debug, info, warn, error |
| Setting | Type | Default | Range | Description |
|---|
codebuddy.hybridSearch.vectorWeight | number | 0.7 | 0–1 | Weight for semantic similarity scores |
codebuddy.hybridSearch.textWeight | number | 0.3 | 0–1 | Weight for BM25 keyword scores |
codebuddy.hybridSearch.topK | integer | 10 | 1–50 | Max results returned |
codebuddy.hybridSearch.mmr.enabled | boolean | false | — | Enable MMR diversity re-ranking |
codebuddy.hybridSearch.mmr.lambda | number | 0.7 | 0–1 | MMR trade-off: 0 = max diversity, 1 = max relevance |
codebuddy.hybridSearch.temporalDecay.enabled | boolean | false | — | Boost recently indexed content |
codebuddy.hybridSearch.temporalDecay.halfLifeDays | integer | 30 | 1–365 | Days until a result’s score is halved |
| Setting | Type | Default | Description |
|---|
codebuddy.permissionScope.defaultProfile | enum | "standard" | Default permission profile: restricted (read-only, no terminal), standard (read/write, safe terminal), trusted (full access, auto-approve) |
codebuddy.accessControl.defaultMode | enum | "open" | Access control mode: open (no restrictions), allow (only listed users), deny (block listed users) |
codebuddy.credentialProxy.enabled | boolean | false | Route API calls through a local proxy that injects credentials from OS keychain |
codebuddy.credentialProxy.rateLimits | object | {"anthropic":60,...} | Per-provider rate limits (requests/min). Keys: provider names, values: 1–3600 |
| Setting | Type | Default | Description |
|---|
codebuddy.connectors.states | object | {} | Connection status of external connectors |
codebuddy.mcp.servers | object | {} | MCP server configurations |
codebuddy.mcp.disabledTools | object | {} | Per-server lists of disabled MCP tool names |
| Setting | Type | Default | Description |
|---|
codebuddy.automations.dailyStandup.enabled | boolean | true | Daily Standup automation (8:00 AM) |
codebuddy.automations.codeHealth.enabled | boolean | true | Daily Code Health Check (9:00 AM) |
codebuddy.automations.codeHealth.hotspotMinChanges | number | 3 | Min changes in 30 days to flag a file as a hotspot |
codebuddy.automations.codeHealth.largeFileThreshold | number | 300 | Min lines for “large file” warning (min: 50) |
codebuddy.automations.codeHealth.maxTodoItems | number | 50 | Max TODO/FIXME/HACK items to report (min: 10) |
codebuddy.automations.dependencyCheck.enabled | boolean | true | Daily Dependency Check (11:00 AM) |
codebuddy.automations.gitWatchdog.enabled | boolean | true | Git Watchdog (every 2 hours) |
codebuddy.automations.gitWatchdog.protectedBranches | string[] | ["main","master","develop",...] | Branches to protect from stale cleanup. Exact names or prefix/* patterns |
codebuddy.automations.endOfDaySummary.enabled | boolean | true | End-of-Day Summary (5:00 PM) |
| Setting | Type | Default | Description |
|---|
codebuddy.failover.enabled | boolean | true | Auto-switch to fallback LLM on failure (rate limit, timeout, auth error) |
codebuddy.failover.providers | string[] | [] | Ordered fallback providers: Anthropic, OpenAI, Gemini, Groq, Deepseek, Qwen, GLM, Local. Empty = auto-detect |
| Setting | Type | Default | Description |
|---|
codebuddy.telemetry.persistTraces | boolean | true | Persist agent telemetry traces to disk |
codebuddy.telemetry.retentionDays | number | 7 | Days to retain traces before pruning (1–90) |
codebuddy.telemetry.otlpEndpoint | string | "" | Generic OTLP HTTP endpoint for trace export (LangSmith, Jaeger, …). Empty = disabled |
codebuddy.telemetry.langfuse.enabled | boolean | false | Enable Langfuse LLM observability. Needs publicKey + secret key (set via command palette) |
codebuddy.telemetry.langfuse.baseUrl | string | "https://cloud.langfuse.com" | Langfuse host URL (cloud or self-hosted) |
codebuddy.telemetry.langfuse.publicKey | string | "" | Langfuse public key (pk-lf-…) |
codebuddy.telemetry.langfuse.allowInsecure | boolean | false | Allow non-HTTPS Langfuse endpoints (self-hosted only) |
codebuddy.telemetry.anonymousUserId | boolean | false | Emit an anonymous per-install id with traces (opt-in; sha256 of a local nonce) |
| Setting | Type | Default | Description |
|---|
codebuddy.multiModel.enabled | boolean | false | Route subagent tasks to the best available model per task category |
codebuddy.multiModel.routing | object | {} | Per-category model overrides |
| Setting | Type | Default | Description |
|---|
codebuddy.quickjs.enabled | boolean | false | Enable the QuickJS runtime for user-authored skill scripts / tool-output transforms (string-io only). See Sandbox |
codebuddy.experimental.codeInterpreter | boolean | false | Experimental: the LLM-callable QuickJS interpreter for recursive/dynamic-subagent workflows |
codebuddy.experimental.codeInterpreterTimeoutSeconds | number | 300 | Max seconds a single eval() may run (clamped 30–600) |
| Setting | Type | Default | Description |
|---|
codebuddy.standup.myName | string | "" | Your name in standup/meeting notes (filters your action items). Falls back to git config user.name |
codebuddy.standup.previewBeforeSave | boolean | true | Preview the generated standup before it’s saved |