Letta MCP Server

remote-capable server

The server can be hosted and run remotely because it primarily relies on remote services or has no dependency on the local environment.

Integrations

  • Uses .env files for configuration, allowing storage of Letta API credentials and endpoints securely outside of the codebase.

  • Repository hosting for the Letta MCP server, enabling version control, contribution workflows, and distribution of the server code.

  • Provides package management for the Letta MCP server, managing dependencies and offering scripts for building and running the server.

Letta MCP Server

A server that provides tools for agent management, memory operations, and integration with the Letta system.

Quick Setup

Option 1: Run with Node.js

# Development (with hot reload) npm run dev:sse # SSE transport # Production npm run build # Build TypeScript first npm run start:sse # SSE transport

Option 2: Run with Docker

# Build and run locally docker build -t letta-mcp-server . docker run -d -p 3001:3001 -e PORT=3001 -e NODE_ENV=production --name letta-mcp letta-mcp-server # Or use the public image docker run -d -p 3001:3001 -e PORT=3001 -e NODE_ENV=production --name letta-mcp ghcr.io/oculairmedia/letta-mcp-server:latest

Directory Structure

  • index.js - Main entry point
  • core/ - Core server functionality
  • tools/ - Individual tool implementations
  • transports/ - Server transport implementations (stdio and SSE)

Available Tools

Agent Management

ToolDescriptionRequired ParametersOptional Parameters
create_agentCreate a new Letta agentname, descriptionmodel, embedding
list_agentsList all available agents-filter
prompt_agentSend a message to an agentagent_id, message-
get_agentGet agent details by IDagent_id-
modify_agentUpdate an existing agentagent_id, update_data-
delete_agentDelete an agentagent_id-
clone_agentClone an existing agentsource_agent_id, new_agent_nameoverride_existing_tools, project_id
bulk_delete_agentsDelete multiple agents-agent_ids, agent_name_filter, agent_tag_filter

Memory Management

ToolDescriptionRequired ParametersOptional Parameters
list_memory_blocksList all memory blocks-filter, agent_id, page, pageSize, label
create_memory_blockCreate a new memory blockname, label, valueagent_id, metadata
read_memory_blockRead a memory blockblock_idagent_id
update_memory_blockUpdate a memory blockblock_idvalue, metadata, agent_id
attach_memory_blockAttach memory to an agentblock_id, agent_idlabel

Tool Management

ToolDescriptionRequired ParametersOptional Parameters
list_toolsList all available tools-filter, page, pageSize
list_agent_toolsList tools for a specific agentagent_id-
attach_toolAttach tools to an agentagent_idtool_id, tool_ids, tool_names
upload_toolUpload a new toolname, description, source_codecategory, agent_id
bulk_attach_tool_to_agentsAttach a tool to multiple agentstool_idagent_name_filter, agent_tag_filter

Additional Tools

  • Model Management: list_llm_models, list_embedding_models
  • Archive Management: list_passages, create_passage, modify_passage, delete_passage
  • MCP Server Management: list_mcp_servers, list_mcp_tools_by_server
  • Import/Export: export_agent, import_agent

Docker Operations

# View container logs docker logs -f letta-mcp # Stop the container docker stop letta-mcp # Update to latest version docker pull ghcr.io/oculairmedia/letta-mcp-server:latest docker stop letta-mcp docker rm letta-mcp docker run -d -p 3001:3001 -e PORT=3001 -e NODE_ENV=production --name letta-mcp ghcr.io/oculairmedia/letta-mcp-server:latest

Configuration with MCP Settings

Add the server to your mcp_settings.json:

"letta": { "command": "node", "args": [ "--no-warnings", "--experimental-modules", "path/to/letta-server/index.js" ], "env": { "LETTA_BASE_URL": "https://your-letta-instance.com", "LETTA_PASSWORD": "yourPassword" }, "disabled": false, "alwaysAllow": [ "upload_tool", "attach_tool", "list_agents", "list_memory_blocks" ], "timeout": 300 }

For remote instances, use the URL configuration:

"remote_letta_tools": { "url": "http://your-server:3001/sse", "disabled": false, "alwaysAllow": [ "attach_tool", "list_agents", "list_tools", "get_agent" ], "timeout": 120 }

You must be authenticated.

A
security – no known vulnerabilities
A
license - permissive license
A
quality - confirmed to work

An MCP server implementation that enables interaction with the Letta API for managing agents, memory blocks, and tools in the Letta system.

  1. Quick Setup
    1. Option 1: Run with Node.js
    2. Option 2: Run with Docker
  2. Directory Structure
    1. Available Tools
      1. Agent Management
      2. Memory Management
      3. Tool Management
      4. Additional Tools
    2. Docker Operations
      1. Configuration with MCP Settings
        ID: fuuaesngfg