Rabbitrabbit/ide
FeaturesToken EconomyModelsAgentPrivacyManifestoPricingDocsBlogChangelogFAQDownload

Docs / Agent

MCP tools reference.

The Rabbit agent registers 45 native tools at startup. All tools are available to every MCP-compatible participant. External MCP servers added via ~/.rabbit-ide/mcp-servers.json extend this list dynamically.

Read & navigate

  • read_file -- cached file read with modes: full, signatures, map, lines
  • search -- workspace-wide regex search with smart-case
  • codebase_search -- semantic vector search over the indexed workspace
  • find_files -- glob-based file discovery
  • tree -- directory tree with token estimates
  • symbols -- file symbol outline via DocumentSymbolProvider
  • find_references -- LSP-precise reference lookup

Edit

  • write_file -- atomic diffs queued per agent turn
  • fast_apply -- local 7B model instant edit application
  • refactor -- cross-project find-and-rename with preview/apply

Shell & tasks

  • run_command -- shell execution with confirmation + compressed output
  • managed_terminal -- persistent agent terminal with output capture
  • multi_shell -- parallel command execution (up to 10)
  • autofix -- run build/test, parse errors, read source context
  • task -- background long-running job management
  • ssh_pool -- pooled remote SSH execution

Project intelligence

  • graph -- knowledge graph with related/impact/path/export actions
  • impact -- file dependents + exported symbols analysis
  • project_profile -- languages, frameworks, entry points scan
  • diff_review -- git diff risk analysis + coverage gaps
  • discover_context -- auto-pull relevant files for a task

Context economy (stoken)

  • intent -- classify task + recommend optimal read plan
  • fill_budget -- pack candidate files into a token budget
  • pack_context -- assemble a session context bundle
  • compact_suggest -- advise when/what to evict from cache
  • analyze_compression -- entropy + best-strategy advisor
  • benchmark_compression -- savings per compression strategy
  • cross_file_dedup -- reclaim duplicated content across files
  • compress_response -- strip filler from long LLM replies
  • file_delta -- token-cheap incremental diffs

Agent & memory

  • spawn_subagent -- cheap-model exploration sub-agent
  • update_plan -- live TODO checklist rendered in chat
  • remember -- persist durable facts to memory store
  • session_memory -- cross-session recall and context
  • use_skill -- load a SKILL.md instruction package
  • skill -- versioned skill registry CRUD

Web & preview

  • web_search -- DuckDuckGo search, no API key required
  • fetch_url -- download + strip HTML page to text
  • open_browser -- open URL in IDE embedded browser preview
  • watch -- tail a file or log for changes

Telemetry & tooling

  • metrics -- session token usage + CEP compliance score
  • wrapped -- token-savings wrapped report (day/week/month/all)
  • semantic -- build / query the local vector index
  • snippet -- save / search reusable code snippets
  • events -- event-bus hook management
  • describe_tool -- on-demand full tool schema lookup

Adding external MCP servers

// ~/.rabbit-ide/mcp-servers.json
{
  "stoken":    { "command": "stoken",    "args": ["serve"] },
  "postgres":  { "command": "mcp-pg",    "args": ["--url", "$DATABASE_URL"] },
  "puppeteer": { "command": "mcp-pptr",  "args": [] },
  "figma":     { "command": "mcp-figma", "env": { "FIGMA_TOKEN": "$FIGMA_TOKEN" } }
}

Tools from external servers appear in the agent prompt automatically. Use describe_tool to inspect any tool's full schema at runtime.