Letta Code CLI
Letta Code runs a coding agent with self-rewriting memory in git, plus skills, subagents, schedules and chat channels
- 3,827 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The agent rewrites its own context and runs hooks and schedules
- Works with files and commands under the chosen permission modes
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/letta-codeDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent rewrites its own context and runs hooks and schedules
- Works with files and commands under the chosen permission modes
This is third-party code. Review the repository files before installing.
What it does
Letta Code is a harness for stateful agents: they have memory, identity and experience that builds up over time. The agent rewrites its own context to adapt to you and the project. You work with it from a local CLI, from the browser including mobile, and through messengers such as Telegram, Slack and Discord. All context, including memory blocks, is tracked in git (MemFS) and synced to your repository with `/memory-repository set`. It loads global skills from `~/.letta`, project skills from `.agents/skills` and the agent's own skills. Built-in subagents, hooks, permission modes, schedules and search across all messages via `/search` are included.
Who it is for. Developers who want a coding agent with long-term memory and scheduled work instead of sessions that start from zero.
Good fit when
- The agent has to remember code style and project decisions from week to week.
- A task is set from Telegram or Slack and a machine at home carries it out.
- Agent context is kept in your own git repository and moved between machines.
- A background subagent combs through logs while the main one writes code.
- The agent checks service health on a schedule and reports problems.
Not a fit when
- You want a simple assistant with no memory that starts clean every time.
- You need remote computers and secrets without a Letta account: those features require sign-in.
- Alpine, a 32-bit system or Windows ARM64: wheels are not built for them.
Example request
Remember that tests in this project run through `pnpm test:unit`, and every morning check failed CI builds and send a digest to Telegram.Limitations
Remote computers and secrets are available only after signing in to Letta. The agent changes its own context, so its behavior drifts from defaults over time: watch the memory blocks in git. Automatic dreaming is off by default on Windows. Python 3.9 or newer is required, with glibc Linux 2.28, macOS 14 and Windows x64 supported. Do not use `sudo` for `npm install -g`, since CLI self-updates would then stop working.
How to disable. Remove the package with `npm uninstall -g @letta-ai/letta-code` or `uv tool uninstall letta`. The `~/.letta` folder and the memory repository remain and are deleted separately.
FAQ
How do I install it?
Run `npm install -g @letta-ai/letta-code`. An alternative without Node.js is `uv tool install letta` or `pipx install letta`, with Node.js bundled.
Where is the agent's memory stored?
In git, in the MemFS system. The sync repository is set with `/memory-repository set git@github.com:...`.
How do I connect Telegram or Slack?
Channels are described in the Letta Code documentation, in the channels section. You can ask the agent to set them up itself.
Which skills are loaded?
Global ones from `~/.letta`, project ones from `.agents/skills` and agent skills from MemFS. `/skills` lists them and `/skill-creator` creates new ones.
What requires a Letta sign-in?
Remote computers and secrets, which are injected as environment variables with obscured values. Everything else runs locally.
Discussion
No questions or reviews yet.
Related
CLI toolsLetta Skills
letta-ai/skills
44 skills from Letta for Claude Code and Letta Code: GitHub, Slack, Notion, Sentry, documents, a browser and agent setup.
A portable .agent folder with memory, skills and protocols that plugs into many coding agents and keeps its knowledge when you switch tools
A background Letta agent reads Claude Code sessions, builds project memory and whispers hints before each prompt.
Instar runs a persistent agent on top of Claude Code or Codex with memory, scheduled jobs and replies over Telegram.
An open coding agent for the terminal, browser and desktop that works with models from OpenAI and Claude to Ollama
ContextLattice: a local context layer for agents
ContextLattice
A local memory and context layer: a session's goal, decisions and evidence are gathered into a compact packet an agent reopens
README in short
The README presents Letta Code as a harness for agents that are more like people than tools: with memory, identity and self-improvement. Install is one command.