Letta Code is an open-source terminal-based AI coding assistant developed by Letta (formerly MemGPT). Built on top of the Letta agent framework, it leverages stateful agent architectures and persistent memory management to execute multi-step software development workflows directly from the command line. In daily operations, Letta Code provides an interactive terminal user interface (TUI) alongside CLI commands. It executes tasks through a modular system of skills and subagents, coordinating actions like file editing, shell command execution, and codebase analysis. The agent integrates with Model Context Protocol (MCP) servers to interact with external developer tools, environments, and local resources. Unlike traditional stateless CLI assistants that reset context per session, Letta Code is designed around memory-centric agent primitives. It can retain context, user preferences, and repository conventions over time, orchestrating subagents with specialized capabilities rather than treating every prompt as a standalone query. Letta Code is distributed as an open-source project hosted on GitHub, with native binary distributions and Node/npm-based packaging for local installation and developer workflow integration.
Terminal UI & CLI Interface — Provides interactive TUI navigation alongside command-line workflows for inspecting diffs and supervising agent execution. - Stateful Agent Memory — Leverages Letta's core memory framework to persist development context, conventions, and operational state across sessions. - Subagent Orchestration — Supports dispatching focused subagents to handle concurrent or specialized development tasks with dedicated UI tracking. - Modular Skill System — Extensible architecture using dedicated skills (such as testing, MCP access, and proof generation) to extend core agent capabilities. - Model Context Protocol (MCP) Integration — Connects to local and remote MCP servers to integrate external toolchains and developer environments.
Use cases
Use Case: Context-Aware Feature Implementation Developers can assign multi-file refactors or new feature development to Letta Code, allowing the agent to remember project conventions, locate relevant modules, and execute the implementation incrementally.
Use Case: Subagent Task Delegation A developer can run Letta Code to decompose a broad task—such as writing end-to-end tests or auditing dependencies—into targeted subagent operations that run and report back through the TUI.
Use Case: MCP-Driven Tool Integration Engineers can connect local development servers and custom MCP endpoints into Letta Code, enabling the terminal agent to query external database schemas, APIs, and custom tooling while modifying source files.
Who it is for
Letta Code is designed for software engineers and DevOps practitioners who prefer terminal-native workflows and require stateful, persistent AI agent interactions across long-running codebases. It is less suited for developers seeking simple single-turn inline editor completions or fully managed, closed-source cloud IDEs.
Install guide
Installation Install Letta Code globally via npm or download the prebuilt binary from the GitHub releases: bash npm install -g @letta-ai/letta-code(Note: Check the official GitHub repository for native binary releases and alternative package manager configurations.)
First Run Initialize configuration and configure your model provider credentials: bash letta-code --init Set your required API keys (e.g., Anthropic or OpenAI) in your environment: bash export OPENAI_API_KEY="your-api-key" export ANTHROPIC_API_KEY="your-api-key"
Example Session Launch Letta Code in the root of your repository to begin an interactive coding session: bash letta-code Inside the prompt, dispatch a task to the agent: text letta> Review src/api/auth.ts, fix the token expiration handling, and run the test suite to verify.
Vestige — Vestige is an open-source agent harness and persistent memory infrastructure framework developed by samvallad33. Designed specifically for AI coding agents…
Beads — Beads (bd) is an open-source task coordination and state orchestration harness for AI coding agents developed by Gastown Hall. It…
GoodMemory — GoodMemory is an open-source agent memory orchestration harness and context management system developed by hjqcan. Designed to give AI agents…
pi-mem — pi-mem is an open-source memory harness and persistent context manager developed by Jo Inc for the Pi CLI coding agent.…
SWE-agent — SWE-agent is an open-source autonomous software engineering agent harness built and maintained by researchers from Princeton University and Stanford University.…
OpenHands — OpenHands (formerly OpenDevin) is an open-source autonomous AI software development agent and orchestration platform. It provides an extensible runtime environment…