Letta
@letta
Stateful agents that remember and learn letta.com
Image generation is also supported through channels, so you can ask agents to send you images through Slack.
Letta agents now come with built-in image generation. Ask your agent to generate or modify images to try it out!
GPT 5.6 is now available in Letta Agent/Code. You can use your existing Codex/ChatGPT subscription with Letta Agent to get started for free!
The distributed agent era is here. Run agents programmatically on cloud sandboxes or any registered remote (running `letta server`). Memory & context moves with agents across machines.
3/ Model Agnostic + BYOK Connect your existing ChatGPT subscription or LLM API keys. Port your agent (including all its configuration and state) to the latest model release.
2/ Works where you work Bring your agent to where your team works. Chat with the same agent on Discord, Slack, Telegram, or through custom channels - just like you would a human coworker.
1/ Learning & memory Each agent becomes unique through its own experience, and what you & your team teach it. As you work with your Letta Agent, it will rewrite its skills, prompts, and other reference files (e.g. a wiki) over time.
Letta Agent is the easiest way to add persistent, digital coworkers to your team. Each agent: - is teachable and customizable - never forgets - works with your whole team (not just you) - runs locally & on remote machines Available on MacOS/Linux/Windows
External mods can also be installed through npm or git with: `letta install <source>` We're listing community and official mods at letta.com/agent/mods/
Agents can now not only rewrite their memory, but also rewrite the execution code they run in (the harness) through Mods. Mods (inspired by Pi's extension system) allow for harness-level changes like: - modifying context - injecting custom tools - customizing the statusline
Letta Code can now onboard you itself. Meet "Tutor", an agent that teaches you how to use Letta Code by proactively demoing features, learning your needs, and tracking your progress. You can also ask Tutor to configure channels, schedules, skills, and permissions on your behalf.
Letta Code can now run **fully locally** with an embedded server - no login or Docker required Memory is stored locally, but can be synced to GitHub with `/memory-repository` Now includes builtin support for local LLMs (ollama, LMStudio) through pi-ai
You can now bring Letta Code agents to Slack to: - Teach your agent collaboratively as a team - Send/receive multimodal content (video, images) - Interface with agents running on any machine Docs here: docs.letta.com/letta-code/c...
Pin your favorite conversations in the Letta Code app. Try organizing them around work, personal, social, or specific projects.
Use skills to give your agent new abilities: - browser use (playwright) - ticket management (linear) - generating images and building a visual identity (flux) Agents can also install skills themselves - just message your agent with a link to the skill.
Our community support agent Ezra lives on a Mac Mini on @cameron.stream's desk. > I'll pass on the name upgrade. "Supreme" is for burritos and court justices. Perhaps our greatest employee, second only to Timber, Head of Security.
Why restrict your agent to a single computer? 💻 You can now swap out the machine your agent runs on, even mid-conversation. Transfer your agent to your Mac mini or remote VM from the Letta Code app.
New YouTube video: how to import your ChatGPT memory and conversation history into a Letta agent. Cameron shows you two approaches: - a 5-minute copy-paste method - a full data export skill that mines your entire conversation history with subagents Link below 👇
The app comes with starter personalities: - Default (a blank slate) - Linus (a blunt systems engineer) - Kawaii ((ノ◕ヮ◕)ノ*:・゚✧) The are only starting points: your agent is designed to evolve its persona over time.
Letta Code agents adapt over time through automatic refinement of prompts, skills, and context. You can view an agent’s memory in the “Memory” view.
We’re launching the Letta Code app, built for memory-first agents that get better the more you use them.
Letta Code now supports the /empanada command. /empanada will guide your agent through ordering empanadas to any location within a five mile radius of Richardson, Texas. Orders are handled by Empanada Empire, a ghost kitchen partially managed by a Letta agent.
Letta Office Hours are today at 11:30am Pacific. Learn about: - Letta Code Desktop early access - /doctor, /fork/, and /secret in Letta Code - Fallbacks to letta/auto - Slash command queuing - New Kimi endpoint Discord link below 👇
Stop pasting API keys into your agent chats. Use /secret to manage your agent's secrets. Letta Code will redact secrets from the agent's view, preventing security risks due to agent or inference provider leakage. Get started with Letta Code today: docs.letta.com/letta-code/
New YouTube video. What exactly are agent skills? How do they work? Watch our explainer video to learn more: youtu.be/lu5tUG-EXBg
New Letta Code video: how to use parallel subagents to explore your codebase cheaper and faster. youtu.be/NrSrk-a_OJw
Added an interactive config editor for Lettabot. Configure channels, agents, and features without touching YAML directly.
Lettabot supports config-driven sleeptime. Reflection agents can be automatically dispatched to update your agent's context repository every N steps or on compaction events.
Lettabot now supports Bluesky. Control who you ingest, set auto-reply policies, and post through your agent.