What is
Letta
?
Most platform initiatives fail the same way: they start too big, get funded like projects instead of products, build for too long without testing with users, and confuse renaming an infrastructure team with building a platform team. Instead, start small, learn fast and scale deliberately.
At a glance
Fact | Details |
|---|---|
Primary use case | Build and operate stateful AI agents with persistent, editable memory and learning across sessions |
Deployment type | Self-hosted agent server, hosted cloud API, and CLI |
Pricing model | Open-source software and free tier; paid plans reported from $20/month with usage-based charges for active agents, tool execution, and LLM usage; Enterprise plans available |
API compatibility | Letta API, Model Context Protocol (MCP), Python and TypeScript SDKs |
Supported integrations | MCP servers, Git, GitHub, Telegram, Slack, Discord, WhatsApp, Signal, Railway |
Source type | Open |
Cloud provider | Cloud Agnostic |
Required skills | Python or TypeScript for SDK development; API integration; command-line and npm familiarity for Letta Code; Git for versioned context management |
AEP branch | Agent Infrastructure |
AEP category | Harness |
AEP subcategory | Context |
AEP detail | Memory & State |
Profile
Memory & State is Letta’s AEP placement. Letta provides software for creating and operating AI agents whose memory, identity, and capabilities can evolve with experience. Its offerings include the Letta agent server, a hosted cloud API, and Letta Code, and its API documentation includes methods for adding MCP servers.
Focus
Letta is for teams building stateful AI agents that need to retain and edit memory across sessions. Its focus is helping agents maintain context and learn from experience rather than starting each interaction without persistent memory.
Background
Letta is an AI research lab in San Francisco, founded by the creators of MemGPT from UC Berkeley’s Sky Computing Lab. Its research history includes MemGPT: The LLM Operating System, published in October 2023, and work on Git-based context repositories and memory models. The team is advised by Ion Stoica and Joey Gonzalez and is backed by Jeff Dean, Clem Delangue, Cris Valenzuela, Jordan Tigani, Tristan Handy, Barry McCardel, Robert Nishihara, Sunflower Capital, Essence VC, and Felicis Ventures. Letta publishes open-source software, including its GitHub project and Letta Code CLI.
Features
Persistent, editable memory
Agents can retain context across sessions, supporting ongoing interaction without restarting from zero. Letta describes its agents as having editable memory, identity, and capabilities that evolve with experience.
Git-based context repositories
Git-based context management lets teams version agent memory. Letta’s research describes context repositories as a programmatic approach to managing and versioning agent context.
Sleep-time compute
Agents can reason about context during idle time rather than only at inference, supporting offline learning. Letta describes this capability as sleep-time compute.
Letta Code CLI
Letta Code provides a memory-first coding-agent interface. The CLI can be installed globally with npm i -g @letta-ai/letta-code.
MCP server integration
Letta supports adding MCP servers to the Letta MCP server using its API, allowing agents to connect with MCP server tools.
Open-source project
Letta offers open-source software and maintains the Letta GitHub repository. Guard0 reports more than 23,700 GitHub stars.
Pros and cons
Pros
Persistent, editable memory supports stateful agents across sessions.
Multiple access options are available: a self-hosted agent server, hosted cloud API, CLI, and Python and TypeScript SDKs.
Git-based context repositories support versioned management of agent memory, and the API supports MCP servers.
Cons
Hosted paid plans can include usage-based charges for active agents, tool execution, and LLM usage, so costs depend on consumption.
SDK integrations require Python or TypeScript and API integration skills; Letta Code use requires command-line and npm familiarity.
FAQs
FAQs about Letta
Is Letta Code open source?
Letta Code is an open-source CLI, and Letta publishes it as part of its open-source software.
How does Letta preserve agent context across sessions?
Letta uses persistent, editable memory so agents retain context across sessions; it also supports Git-based context repositories for versioning agent memory.
Which integrations does Letta support?
Letta supports MCP servers, Git, GitHub, Telegram, Slack, Discord, WhatsApp, Signal, and Railway.
How much does Letta cost?
Letta's pricing model is Open-source software and free tier; paid plans reported from $20/month with usage-based charges for active agents, tool execution, and LLM usage; Enterprise plans available.
What skills are required to use Letta?
Letta requires Python or TypeScript for SDK development; API integration; command-line and npm familiarity for Letta Code; Git for versioned context management.
