Letta MCP Server Railway Edition

You deploy and run a cloud-optimized MCP server that connects AI clients to Letta.ai’s stateful agents over streamable HTTP. This edition is designed for quick cloud deployment, seamless scaling, and reliable client integration across popular MCP clients.

How to use

You connect an MCP client to the server URL to start sending requests for agent management, conversations, and memory tools. The server supports streamable_http transport, which is optimized for cloud deployment and auto-scaling on Railway. Use the provided MCP URL to feed your client configuration, and reference the transport type to ensure compatibility with your MCP client settings. You can test the endpoint with the MCP Inspector or your preferred client to verify connectivity, health, and basic operations like listing agents, sending messages, and querying conversations.

How to install

Prerequisites: Python 3.8+ and a Letta API key from api.letta.com. You also need a Railway account for cloud deployment.

Step-by-step setup locally and for cloud deployment follows.

Steps you can follow

1. Obtain your Letta API key from api.letta.com and keep it handy for environment configuration.

2. For local testing, clone the project and install dependencies, then start the server locally using Python.

Configuration and operation notes

The server runs with environment variables that configure the Letta API access and timeouts. In cloud deployments, Railway manages the port and runtime, while keeping the server accessible at the mcp path. You can adjust timeouts and retry behavior via environment variables to suit your workload.

Troubleshooting

If you encounter connection issues, verify that your Railway app is running and that the MCP URL is reachable. Check the health endpoint and ensure the LETTA_API_KEY is correctly set in your environment. For timeouts, increase the client-side timeout in your MCP configuration.

Available tools

letta_list_agents

List all agents with pagination and filtering

letta_create_agent

Create new agents with memory blocks and tools

letta_get_agent

Get detailed agent information

letta_update_agent

Update agent configuration (name, description, model)

letta_delete_agent

Safely delete agents with confirmation

letta_send_message

Send messages to agents with streaming support

letta_get_conversation_history

Retrieve chat history with pagination

letta_export_conversation

Export conversations (markdown, JSON, text)

letta_get_memory

View all memory blocks for an agent

letta_update_memory

Update memory blocks (human, persona, custom)

letta_create_memory_block

Create custom memory blocks

letta_search_memory

Search through agent conversation memory

letta_list_tools

List all available tools with filtering

letta_get_agent_tools

View tools attached to specific agents

letta_attach_tool

Add tools to agents

letta_detach_tool

Remove tools from agents

letta_health_check

Verify API connection and service status

letta_get_usage_stats

Get usage statistics and analytics