Letta Code | Letta Docs

# Letta > Letta is an open source framework for building stateful AI agents with persistent memory, tool execution, and multi-agent coordination. ## Complete Documentation Download For the full documentation in a single file (guides, tutorials, concepts): - **llms-full.txt**: https://docs.letta.com/llms-full.txt This ~2MB file is ideal for: - Uploading to Claude, ChatGPT, or other LLMs - Local reference without scraping - Building custom AI assistants with Letta knowledge Note: API reference is not included in llms-full.txt (too large). Access specific endpoints via /api/.../index.md URLs. Letta enables developers to create AI agents that remember context across conversations, manage their own memory through tool calls, and perform complex multi-step tasks. The framework provides a REST API and SDKs (Python, TypeScript) for building production agent applications. Key capabilities: - Persistent memory (in-context blocks + archival/RAG) - Custom tool execution (sandbox, client-side, MCP) - Multi-agent coordination patterns - Streaming responses - Human-in-the-loop workflows ## SDK Quick Reference ### TypeScript ```typescript import { LettaClient } from "@letta-ai/letta-client"; // Initialize client (Cloud) const letta = new LettaClient({ apiKey: "...", projectId: "..." }); // Create agent const agent = await letta.agents.create({ name: "my-agent", model: "openai/gpt-4o", embedding: "openai/text-embedding-3-small" }); // Send message const response = await letta.agents.messages.create(agent.id, { messages: [{ role: "user", content: "Hello!" }] }); // Memory blocks await letta.agents.blocks.create(agent.id, { label: "notes", value: "..." }); await letta.agents.blocks.update(agent.id, blockId, { value: "updated" }); // Custom tools const tool = await letta.tools.create({ source_code: "def greet(name: str):\n return f'Hello {name}'" }); await letta.agents.tools.attach(agent.id, tool.id); // Folders (document storage) const folder = await letta.f