A TypeScript coding agent CLI where a single 'ulw' or 'mass ulw' keyword in your prompt makes it plan, delegate to multiple models, and verify the work.
letta-ai/letta-code
A stateful agent harness for building AI agents with persistent memory, identity, and skills that learn over time, usable from a CLI, browser, or chat apps.
- Stars
- 3,677
- Forks
- 443
- Open issues
- 515
- Last commit
- 2 hours ago
What is it?
Letta Code is a harness for running stateful AI agents that keep memory, a persistent identity, and a record of past experience across sessions. Agents can rewrite their own memory blocks, skills, and prompts to adapt over long periods, and they can be configured to run periodic background consolidation.
You can talk to an agent from a local CLI, from the browser at chat.letta.com, or through messaging integrations such as Telegram, Slack, and Discord. Context is tracked in git through MemFS, so it can be synced to a custom GitHub repository. Agents can also search past messages and call other agents, including themselves, as subagents.
The tool supports skills loaded from global, project, and agent scopes, hooks for running scripts at key execution points, and permission modes for controlling auto-approved actions. Scheduling with heartbeats and crons lets agents work proactively, and remote computers and secrets can be used across machines when signed in to Letta.
Key features
- 01Persistent memory blocks and git-tracked context through MemFS
- 02Skills loaded from global, project, and agent scopes
- 03Subagents and agent-to-agent calls, including background execution
- 04Messaging channels for Slack, Telegram, Discord, and the browser
- 05Hooks, permission modes, and crons for automation
Use cases
- Building a long-running personal assistant that remembers past conversations
- Running proactive agents on a schedule with heartbeats and crons
- Chatting with the same agent from a laptop, phone, or Slack
Quick start
- 1
Install the CLI via npm
npm install -g @letta-ai/letta-code - 2
Navigate to your project directory and run the CLI
letta - 3
Optionally run the tutorial agent
letta --new-agent --personality tutorial
Things to know
- Remote computers, secrets, and some features require signing in with Letta.
- Automatic client-side dreaming is disabled by default on native Windows.
- Native Windows ARM64, Alpine/musl Linux, and 32-bit systems are unsupported by the Python wheels.
Topics
FAQ
- What is letta-code?
- A stateful agent harness for building AI agents with persistent memory, identity, and skills that learn over time, usable from a CLI, browser, or chat apps. Letta Code is a harness for running stateful AI agents that keep memory, a persistent identity, and a record of past experience across sessions. Agents can rewrite their own memory blocks, skills, and prompts to adapt over long periods, and they can be configured to run periodic background consolidation.
- Is letta-code free to use? What license does it have?
- letta-code is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is letta-code still maintained?
- Actively maintained. The last commit to letta-code was on Oct 22, 2026.
- What can I use letta-code for?
- Building a long-running personal assistant that remembers past conversations Running proactive agents on a schedule with heartbeats and crons Chatting with the same agent from a laptop, phone, or Slack
- How do I install letta-code?
- Install the CLI via npm: npm install -g @letta-ai/letta-code Navigate to your project directory and run the CLI: letta Optionally run the tutorial agent: letta --new-agent --personality tutorial
- Where can I get letta-code?
- The source code and installation instructions are on GitHub: https://github.com/letta-ai/letta-code
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source AI memory platform that turns documents, code and conversations into a self-hosted knowledge graph for AI agents.
A personal context store that gives AI coding agents like Codex, Claude Code and OpenCode persistent project knowledge via a CLI, MCP server and desktop app.
A workflow and runtime layer for OpenAI Codex CLI that adds structured planning, durable goal execution, parallel agent teams, and state tracking under .omx/.
A CLI that scans your codebase and generates compact AI context files (CLAUDE.md, .cursorrules, AGENTS.md) so coding assistants spend fewer tokens exploring the project.
On-screen AI assistant for desktop that reads the current app's context and uses multiple LLMs and MCP tools to help you act across apps.