Letta Agent

What it does

Letta Agent is a platform for building stateful agents that keep their own identity and long-term memory across conversations, models, and devices. You configure an agent with a persona, skills, tools, and scheduled tasks, then interact with it through the browser, a terminal CLI, or the Letta Agent SDK. Agents can also be reached through Slack, Discord, Telegram, or custom channels.

People use Letta Agent to create coding assistants, personal work assistants, and AI coworkers that improve with use. Its distinct features include memory that is git-tracked through MemFS, sleep-time compute for self-improvement without model updates, and the ability to connect your own device or VM as a remote environment. Agents can move between models without losing memory, and you can bring your own API keys or existing coding plans.

Letta Agent suits developers and teams building assistants that maintain context and memory across sessions. You can connect your own model API keys or a supported coding subscription, with provider charges and limits still applying. The CLI and SDK support different cloud, local and remote execution arrangements. Teams should review tool permissions, credentials and channel setup for the chosen backend, and test how memory is updated and reused before entrusting an agent with important work.

What it can do

• Create persistent custom agents with their own persona, skills, harness modifications, and tools.
• Self-improve through sleep-time compute, with context such as skills and prompts git-tracked through MemFS.
• Communicate with agents through Slack, Discord, Telegram, or custom channels.
• Schedule proactive tasks, let agents set their own schedules, and review past runs.
• Connect your own device, sandbox, or VM as a remote environment, with agents working across multiple devices.
• Switch models while keeping agent memory, using OpenAI Codex, Z.AI, other subscription plans, or your own API keys.

Works with

• Slack
• Discord
• Telegram
• GitHub

Example workflows

Persistent research analyst

Illustrative example. Input: a persona such as a competitor-tracking analyst, created via the SDK. Steps: create the agent once, then resume the session later and send a follow-up question. Output: the agent responds using the memory it retained from earlier conversations.

Scheduled team reporting in chat

Illustrative example. Input: a recurring task and a Slack or Telegram channel connection. Steps: set a schedule for the agent, let it run proactively, and review past runs. Output: periodic results delivered and reviewable in the connected channel.

Before you choose

• Bring eligible model access and account for provider charges and limits.
• Configure tool permissions and credentials for the selected cloud, local or remote backend.
• Review remembered information and test channel behavior before deploying an agent.

Listed under

• Coding
• Coding agents
• AI agents
• Developers
• Persistent memory
• Scheduled tasks
• Team agents

Get started

Start with the Letta web app or CLI, or build with the Letta Agent SDK.

1. Install the Letta CLI or the Letta Agent SDK, following the quickstart.
2. Connect your own API keys or a supported coding plan to the agent.
3. Configure channels such as Slack or Discord, or set up a local-backend channel server if you run one yourself.

Connecting your own API keys or supported coding plans to Letta Agent is free. Model-provider charges, subscription eligibility and usage limits remain separate. Confirm any hosting or enterprise requirements for the selected deployment.

Letta Agent is free to use. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Letta Agent | Personalized Agent That Remembers and Learns ↗www.letta.com
• Letta Documentation | Letta Docs ↗docs.letta.com
• Letta Agent SDK overview | Letta Docs ↗docs.letta.com

Company website

letta.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.