MCP Tools Reference

The main context optimization tool. Returns the most relevant documentation sections for a query, respecting your token budget.

Parameters

Response Format

{
  "sections": [
    {
      "title": "Authentication Flow",
      "content": "...",
      "file": "docs/auth.md",
      "lines": [45, 120],
      "relevance_score": 0.94,
      "token_count": 1200
    }
  ],
  "total_tokens": 3800,
  "suggestions": ["Also check: docs/security.md"]
}

Example

// Query example
{
  "query": "How does user authentication work?",
  "max_tokens": 4000,
  "search_mode": "hybrid"
}

Discover the right MCP tool for your task. Get recommendations based on your query, detailed tool info, or browse tools by tier. Perfect for new users or when you're unsure which tool to use.

Parameters

Response Format

{
  "recommendations": [
    {
      "tool": "rlm_multi_project_query",
      "score": 92,
      "tier": "TEAM",
      "description": "Query across all team projects",
      "use_cases": ["Cross-project search", "Find implementations"],
      "example": "rlm_multi_project_query(query='authentication')"
    }
  ],
  "total_tools": 46,
  "tip": "Use tier='PRIMARY' to see essential tools"
}

Example

// Get tool recommendations
rlm_help({ query: "I want to search across all my team projects" })

// Get info about a specific tool
rlm_help({ tool: "rlm_context_query" })

// List all primary tools
rlm_help({ tier: "PRIMARY" })

Break a complex query into sub-queries with an execution plan. Your LLM can then call rlm_multi_query to get context for each sub-query.

Parameters

Response Format

{
  "original_query": "Explain the full authentication system",
  "subqueries": [
    "How does login flow work?",
    "How are JWT tokens generated?",
    "How are sessions managed?",
    "How does logout work?"
  ],
  "execution_plan": "sequential",
  "estimated_tokens": 12000
}

Execute multiple queries in a single call with per-query token budgets. Efficient for parallel context retrieval.

Parameters

Response Format

{
  "results": [
    {
      "query": "How does login flow work?",
      "sections": [...],
      "token_count": 1800
    },
    {
      "query": "How are JWT tokens generated?",
      "sections": [...],
      "token_count": 1500
    }
  ],
  "total_tokens": 6500
}

Generate a full execution plan for complex questions, including query decomposition, dependencies, and optimal execution order.

Parameters

Response Format

{
  "plan": {
    "steps": [
      { "id": 1, "query": "...", "depends_on": [] },
      { "id": 2, "query": "...", "depends_on": [1] }
    ],
    "parallel_groups": [[1, 2], [3]],
    "estimated_tokens": 15000
  }
}

Query across all projects in a team with a single call. Returns context from multiple projects ranked by relevance. Requires a team API key.

Parameters

Response Format

{
  "query": "How does authentication work?",
  "projects_queried": 5,
  "projects_skipped": 0,
  "results": [
    {
      "project_slug": "api-service",
      "sections": [...],
      "tokens": 800
    },
    {
      "project_slug": "web-app",
      "sections": [...],
      "tokens": 600
    }
  ],
  "total_tokens": 3200
}

Legacy query tool using keyword search. Use rlm_context_query instead for better results.

Parameters

Response Format

{
  "content": "Relevant documentation content...",
  "sources": ["docs/guide.md:45-120"]
}

Search documentation using regex patterns. Useful for finding specific function names, configuration keys, or code patterns.

Parameters

Response Format

{
  "matches": [
    {
      "file": "docs/config.md",
      "line": 45,
      "content": "API_KEY=your-key-here",
      "context": "..."
    }
  ],
  "total_matches": 15
}

Inject session context that persists across queries. Useful for setting task-specific focus areas.

Parameters

Response Format

{
  "success": true,
  "context": "Task: Fix authentication bug. Focus: auth.ts, middleware.ts"
}

Show the current session context. Useful for checking what context is active.

Parameters

Response Format

{
  "context": "Task: Fix authentication bug. Focus: auth.ts, middleware.ts",
  "injected_at": "2024-01-15T10:30:00Z"
}

Clear the session context. Use when switching tasks.

Parameters

Response Format

{
  "success": true,
  "message": "Session context cleared"
}

Get statistics about the indexed documentation.

Parameters

Response Format

{
  "total_documents": 42,
  "total_sections": 156,
  "total_tokens": 125000,
  "last_updated": "2024-01-15T10:30:00Z"
}

List all documentation sections. Useful for understanding what topics are covered.

Parameters

Response Format

{
  "sections": [
    { "id": "auth-overview", "title": "Authentication Overview", "file": "docs/auth.md" },
    { "id": "api-keys", "title": "API Keys", "file": "docs/auth.md" }
  ]
}

Read specific line ranges from documentation. Useful after searching to get full context.

Parameters

Response Format

{
  "content": "Full content of lines 100-150...",
  "file": "docs/auth.md",
  "lines": [100, 150]
}

Get project settings from the Snipara dashboard. Returns configuration like auto-inject preferences, default search mode, etc.

Parameters

Response Format

{
  "projectId": "abc123",
  "name": "My Project",
  "defaultSearchMode": "hybrid",
  "autoInjectInstructions": true,
  "maxTokensDefault": 4000
}

Store an LLM-generated summary for a document. Later queries can use stored summaries for faster, more token-efficient responses.

Parameters

Response Format

{
  "summary_id": "sum_abc123",
  "document_path": "docs/auth.md",
  "summary_type": "concise",
  "token_count": 150,
  "created": true,
  "message": "Summary stored successfully"
}

Retrieve stored summaries with optional filters. Use to check what summaries exist or to use them in context.

Parameters

Response Format

{
  "summaries": [
    {
      "summary_id": "sum_abc123",
      "document_path": "docs/auth.md",
      "summary_type": "concise",
      "token_count": 150,
      "content": "Authentication uses JWT tokens...",
      "created_at": "2024-01-15T10:30:00Z"
    }
  ],
  "total_count": 1,
  "total_tokens": 150
}

Delete stored summaries by ID, document path, or type.

Parameters

Response Format

{
  "deleted_count": 3,
  "message": "Deleted 3 summaries"
}

Get merged context from linked shared collections. Returns team coding standards and best practices with token budget allocation by category priority.

Parameters

Response Format

{
  "documents": [
    {
      "id": "doc_123",
      "title": "TypeScript Standards",
      "category": "MANDATORY",
      "token_count": 800,
      "collection_name": "Team Coding Standards"
    }
  ],
  "merged_content": "# TypeScript Standards\n...",
  "total_tokens": 2400,
  "collections_loaded": 2
}

List available prompt templates from linked shared collections. Templates are reusable prompts for common tasks.

Parameters

Response Format

{
  "templates": [
    {
      "id": "tpl_123",
      "name": "Security Review",
      "slug": "security-review",
      "description": "Review code for security issues",
      "category": "review",
      "collection_name": "Team Templates"
    }
  ],
  "total_count": 5,
  "categories": ["review", "refactoring", "testing"]
}

Get a specific prompt template and optionally render it with variable substitution.

Parameters

Response Format

{
  "template": {
    "id": "tpl_123",
    "name": "Security Review",
    "slug": "security-review",
    "prompt": "Review the following code for {{focus_area}}:\n{{code}}",
    "variables": ["focus_area", "code"]
  },
  "rendered_prompt": "Review the following code for SQL injection:\n...",
  "missing_variables": []
}

List all shared context collections accessible to you. Returns collections you own, team collections you're a member of, and public collections. Use this to discover collection IDs for uploading documents.

Parameters

Response Format

{
  "collections": [
    {
      "id": "col_abc123",
      "name": "Team Coding Standards",
      "slug": "team-coding-standards",
      "description": "Shared coding guidelines for all projects",
      "scope": "team",
      "access_type": "team_member",
      "_count": {
        "documents": 12,
        "templates": 5
      }
    },
    {
      "id": "col_def456",
      "name": "Public TypeScript Guide",
      "slug": "public-ts-guide",
      "scope": "public",
      "access_type": "public",
      "_count": {
        "documents": 8,
        "templates": 0
      }
    }
  ],
  "count": 2
}

Example

// List all accessible collections
rlm_list_collections()

// Exclude public collections
rlm_list_collections({ include_public: false })

Upload or update a document in a shared context collection. Use for team best practices, coding standards, and guidelines that should be available across projects.

Parameters

Response Format

{
  "success": true,
  "document_id": "doc_xyz789",
  "collection_id": "col_abc123",
  "title": "Error Handling Standards",
  "category": "BEST_PRACTICES",
  "action": "created"
}

Example

// Upload a new coding standard
rlm_upload_shared_document({
  collection_id: "col_abc123",
  title: "Error Handling Standards",
  content: "# Error Handling\n\nAlways use custom error classes...",
  category: "BEST_PRACTICES",
  priority: 50,
  tags: ["errors", "typescript"]
})

Upload or update a single document. Creates the document if it doesn't exist, updates if it does.

Parameters

Response Format

{
  "success": true,
  "document_id": "doc_abc123",
  "path": "docs/getting-started.md",
  "action": "created",
  "token_count": 1500
}

Bulk sync multiple documents in a single call. Efficient for initial project setup or major updates.

Parameters

Response Format

{
  "success": true,
  "created": 5,
  "updated": 2,
  "deleted": 0,
  "total_documents": 7,
  "total_tokens": 15000
}

Load a specific document's content from your Snipara project into the RLM runtime environment. Returns optimized content ready for REPL injection.

Parameters

Response Format

{
  "success": true,
  "document_path": "docs/api.md",
  "content": "# API Reference\n...",
  "tokens": 3200,
  "sections": 12
}

Load the entire project context into the RLM runtime. Returns a structured overview of all indexed documents with section summaries for broad context injection.

Parameters

Response Format

{
  "success": true,
  "project_slug": "my-project",
  "documents": 15,
  "total_sections": 142,
  "content": "# Project: my-project\n...",
  "tokens": 7500
}

Orchestrate multi-step document operations combining queries, decomposition, and context loading into a single coordinated execution plan.

Parameters

Response Format

{
  "success": true,
  "steps_executed": 4,
  "context_gathered": "# Orchestration Result\n...",
  "tokens_used": 12400,
  "sub_queries": ["auth middleware", "JWT validation", "session management"]
}

Bridge between Snipara MCP and the RLM Runtime REPL. Inject Snipara context into REPL variables or extract REPL state back to Snipara for persistence.

Parameters

Response Format

{
  "success": true,
  "action": "inject",
  "key": "auth_docs",
  "tokens": 2800,
  "repl_session": "default"
}

Recipe 1: Debug an Auth Flow

Find and understand authentication issues in your codebase.

// Step 1: Find auth-related code
rlm_search({ pattern: "authenticate|login|session|jwt" })

// Step 2: Get detailed context on the auth system
rlm_context_query({
  query: "authentication flow and session management",
  max_tokens: 6000,
  search_mode: "hybrid"
})

// Step 3: Check team security standards
rlm_shared_context({ categories: ["MANDATORY", "BEST_PRACTICES"] })

Recipe 2: Generate Feature with Repo Context

Break down a complex feature and get relevant context for each part.

// Step 1: Break down the task into subtasks
rlm_decompose({
  query: "implement password reset with email verification",
  max_subqueries: 5
})

// Step 2: Get context for each subtask in parallel
rlm_multi_query({
  queries: [
    "email service configuration and templates",
    "password hashing and validation",
    "reset token generation and expiry"
  ],
  tokens_per_query: 2000
})

// Step 3: Check existing patterns
rlm_shared_context({ categories: ["BEST_PRACTICES"] })

Recipe 3: Onboard a New Developer

Get a quick overview of a codebase for new team members.

// Step 1: Get project statistics
rlm_stats()

// Step 2: List all documentation sections
rlm_sections()

// Step 3: Get high-level summaries of key areas
rlm_get_summaries({
  summary_type: "overview",
  include_content: true
})

// Step 4: Check team coding standards
rlm_shared_context({
  categories: ["MANDATORY", "BEST_PRACTICES", "GUIDELINES"]
})