Letta
Letta is a San Francisco-based AI research lab, spun out of UC Berkeley, that builds a memory-augmented agent platform enabling AI agents to persist, learn, and self-improve across sessions for developers and enterprise teams.
- Company typePrivate
- Founded2024
- HeadquartersSan Francisco, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What Letta does
Letta is an AI research lab headquartered in San Francisco that builds infrastructure for stateful, self-improving AI agents. Founded in September 2024 by Charles Packer (CEO) and Sarah Wooders (CTO), the creators of the MemGPT research project, the company is a spin-out of UC Berkeley's Sky Computing Lab and is advised by Ion Stoica and Joey Gonzalez. Its core thesis is that current LLM-based agents are fundamentally stateless: Letta's platform supplies persistent memory, context management, and self-improvement primitives that allow agents to retain identity, accumulate learned skills, and recover from errors across sessions.
The product surface spans Letta Cloud (a hosted platform for deploying and managing agent fleets at scale), the Agent Development Environment (ADE) for visual debugging of context and reasoning, the Letta Code memory-first coding agent and desktop application, and a public REST API with auto-generated Python and TypeScript SDKs. Underlying these products is a layered memory architecture including Memory Blocks (structured, tool-editable context), Context Repositories (git-versioned memory), Sleep-time Compute (offline consolidation), Skill Learning (experience-driven skill acquisition), and a suite of evaluation frameworks (Letta Evals, Context-Bench, Recovery-Bench, Letta Leaderboard). The platform is model-agnostic, integrating with OpenAI, Anthropic, Google, and open-source LLMs, and ships with 7,000+ pre-built tools via Composio plus MCP server support.
Letta monetizes through managed Letta Cloud subscriptions ($20-200/month tiers), API/SDK access, and a freemium Letta Code tier, with a product-led growth motion targeting developers and AI engineering teams. Named customers include Bilt (1M+ agent deployment for neighborhood commerce), Kognitos ($500K downstream contract expansion for enterprise analytics), Hunt Club (Hunter AI recruiting agent), and 11x (sales automation Deep Research agent that drove 3 to 85 customers overnight), distributed across loyalty, automation, recruiting, and sales verticals. The company has raised $10 million in seed funding from Felicis, Y Combinator, Sunflower Capital, Essence VC, and angels including Jeff Dean and Clem Delangue.
Letta firmographics
Firmographics- Name
- Letta
- Legal name
- Letta Inc.
- Website
- https://letta.one
- Company type
- Private
- Founded year
- 2024
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- Letta is a San Francisco-based AI research lab, spun out of UC Berkeley, that builds a memory-augmented agent platform enabling AI agents to persist, learn, and self-improve across sessions for developers and enterprise teams.
- Ownership category
- akta.pro rank
Letta industry classification
Industry- Product category
- AI Agent Development Platform
- NAICS
- Software Publishers (5132), Computer Systems Design and Related Services (54151)
- SIC
- Services-Computer Programming Services (7371), Services-Prepackaged Software (7372)
- akta.pro primary industry
- AI Application Enablement Platforms (Copilot/Agent Frameworks, SDKs) (HDAEANAJ)
- akta.pro secondary industries
- API Documentation & Developer Portals (BPAMAOAH), Developer APIs & Node Access (RPC providers, node infrastructure APIs) (FSAPABAC)
Keywords
Where Letta is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Letta business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Infrastructure, Marketing or Sales
Revenue model
- Letta Cloud (Hosted Service): Managed cloud service for deploying and running stateful AI agents with advanced memory capabilities. Developers can build and deploy agents via hosted platform with pricing tiers from $20-200/month.
- Letta Code Premium Tiers: Letta Code desktop application available with free tier for users to get started, with premium tiers for advanced features and usage.
- API Access and SDK Licensing: REST API access and client SDKs (Python, TypeScript) for integrating Letta's agent framework into production applications.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Subscription | Monthly | Letta Cloud hosted service for deploying stateful AI agents |
| Freemium | Monthly | Letta Code desktop application - free tier |
Go-to-market motion2 records
Distribution channels5 records
Marketing channels7 records
Letta product offering
Product offeringCore offering
Letta provides a software platform and developer framework for building, deploying, and managing stateful AI agents with persistent memory. Its core offering centers on Letta Cloud (a hosted agent deployment platform), Letta Code (a memory-first coding agent), and the Agent Development Environment (ADE) — a visual interface for designing and debugging agents. The platform exposes a REST API with Python and TypeScript SDKs, enabling developers to integrate memory-augmented agent capabilities into their own applications.
Product overview
Letta is an AI research lab in San Francisco building machines that learn, offering a comprehensive platform for developing stateful AI agents with persistent memory. The product portfolio centers on Letta Cloud (hosted deployment platform), Letta Code (memory-first coding agent), and the Agent Development Environment (visual debugging interface), supported by a REST API with Python and TypeScript SDKs. Core memory capabilities include Memory Blocks (structured context management), Context Repositories (git-versioned memory), and Sleep-time Compute (offline learning). The platform provides evaluation tools including Letta Evals, Context-Bench, Recovery-Bench, and the Letta Leaderboard for benchmarking agent performance. Letta supports extensibility through Composio (7,000+ tools), MCP connectors, and Skill Learning for dynamic capability acquisition. The company's research foundation traces to MemGPT, an open-source project on virtual context management for LLM agents.
Differentiator
Problem solved
Functional benefit
Brands
- Letta Code: A memory-first coding agent that persists across sessions and improves with use, ranking as the #1 model-agnostic open-source agent on Terminal-Bench.
- Agent Development Environment (ADE)
- Letta Cloud
- Letta Evals
Products and services
- Letta Cloud Hosted platform for developing and deploying stateful AI agents with advanced memory capabilities. Enables developers to build and manage agents via API without managing their own server infrastructure. Pricing tiers range from $20–200/month.
- Letta Code Memory-first coding agent that persists context across sessions and learns from user preferences, code reviews, and action outcomes. Ranks as the #1 model-agnostic open-source harness on Terminal-Bench. Distributed as the @letta-ai/letta-code npm package with a free tier and premium tiers, plus a cross-platform desktop app.
- Agent Development Environment (ADE) Visual development environment that brings transparency to agent design and debugging by making context windows and agent reasoning visible to developers. Provides tools for state management, memory architecture, tool integration, and production deployment simulation.
- Letta REST API and Client SDKs Public REST API for creating and managing stateful AI agents, supporting sync, async, and streaming modes across cloud and self-hosted deployments. Includes auto-generated Python and TypeScript client SDKs with complete feature parity.
- MemGPT Open-source research project on virtual context management and self-editing memory for LLM agents. The original MemGPT framework (pypi package and Docker image) is distributed under an open-source license and serves as the foundation for Letta's agent memory architecture.
Quantifiable outcome
- 42.5% on Terminal-Bench, ranking #1 open-source and #4 overall
- +4 more outcomes
Companies that use Letta
Customer profileNamed customers4 records
Segments3 records
Ideal customer profiles2 records
Letta technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration6 records
AI capability8 records
Feature11 records
Letta partnerships and signals
Strategic signalPartnerships
Three partnerships are on record, tiered core and minor.
- DeepLearning.AIcoreEducational partnership with Andrew Ng's DeepLearning.AI to offer course 'LLMs as Operating Systems: Agent Memory' covering fundamentals of adding memory to LLM agents (~1.5 hours of content).
- ComposiocoreIntegration partnership providing access to 7,000+ pre-built tools through Letta's Agent Development Environment, enabling agents to access extensive tool ecosystem.
- Sundae BarminorSundae Bar's AI agent marketplace (sundaebar.ai) features Letta as the first developer to list an agent, enabling monetization of AI agents.
Scale indicators6 records
Recent moves8 records
Expansion highlights6 records
Letta competitors and assessment
Company assessmentDirect peers
- LangChain: LangChain is the most established open-source framework for building LLM applications and agents, with extensive memory and context management features. It directly competes with Letta in the developer agent framework category, with significantly larger distribution and community.
- Mem0: Mem0 is a dedicated memory layer for AI agents, directly comparable to Letta's memory-focused positioning. Mem0 offers a Pro tier at $249/month, placing it in direct pricing competition with Letta Cloud for the same memory-management use case.
- LlamaIndex: LlamaIndex is an open-source framework for building LLM applications with strong data ingestion, retrieval, and agent capabilities. It overlaps directly with Letta in the agent development framework space, especially for RAG and memory-augmented applications.
- CrewAI: CrewAI is a multi-agent orchestration framework enabling coordinated AI agent teams. It competes with Letta's agent platform for developers building multi-agent systems, though Letta's memory layer is more differentiated.
- AutoGen (Microsoft): Microsoft's AutoGen is an open-source framework for building multi-agent AI applications. Backed by Microsoft research, it competes with Letta's agent framework for enterprise and developer mindshare, with significantly greater distribution through the Microsoft ecosystem.
Emerging players
- Inkeep: Inkeep is an AI agent platform focused on customer support and documentation use cases. It represents an emerging player in the agent framework space with partial overlap to Letta's enterprise targeting, though narrower in vertical focus.
- Relevance AI: Relevance AI is an AI agent platform targeting enterprise teams with a low-code approach to building and deploying agents. It overlaps with Letta's enterprise agent use cases but emphasizes a different (more visual, less developer-native) workflow.
- Fixpoint: Fixpoint is an emerging AI infrastructure company focused on context and memory management for AI applications. It overlaps with Letta's core memory management positioning, though it remains earlier-stage and less proven in production deployments.
Broad incumbents
- OpenAI (with native memory): OpenAI has progressively shipped native memory features in ChatGPT and the Assistants API. As the dominant foundation model provider, OpenAI represents both a key partner (Letta works with OpenAI models) and a competitive threat if native memory becomes 'good enough' to bypass standalone frameworks.
Others
- Pinecone: Pinecone is a vector database used as part of the agent memory stack (used by Letta customers like 11x). It is an enabling/infrastructure peer rather than a direct competitor, but represents an adjacent category that could compete if vector DBs evolve to include more agent memory primitives.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights6 records
Customer concentration
Letta social profiles
Digital presenceLetta compliance and trust
Trust signalCompliance2 records
Letta financial estimates
Financial estimateRevenue estimate
Valuation estimate
Letta leadership team
Management profileNumber of profiles
Profiles4 records
Letta funding detail
Funding detailFunding overview
Funding rounds1 record
Investors4 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Letta M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Letta
What does Letta do?
Letta provides a software platform and developer framework for building, deploying, and managing stateful AI agents with persistent memory. Its core offering centers on Letta Cloud (a hosted agent deployment platform), Letta Code (a memory-first coding agent), and the Agent Development Environment (ADE) — a visual interface for designing and debugging agents. The platform exposes a REST API with Python and TypeScript SDKs, enabling developers to integrate memory-augmented agent capabilities into their own applications.
Is Letta a public or private company?
Letta is a private company. It is classified as venture growth investor backed and is currently operating.
When was Letta founded?
Letta was founded in 2024. It employs 1 to 10 people.
Where is Letta based?
Letta is headquartered in San Francisco, United States, in the North America region.
How does Letta make money?
Three revenue lines are on record. Letta Cloud (Hosted Service) is the primary driver. The others are letta Code Premium Tiers and API Access and SDK Licensing.
Who are Letta's main competitors?
Direct peers on record are LangChain, Mem0, LlamaIndex, CrewAI and AutoGen (Microsoft). Emerging players are Inkeep, Relevance AI and Fixpoint. OpenAI (with native memory) is listed as a broad incumbent. Pinecone is listed as an others.
Does Letta have an API?
Yes. Letta offers a REST API for building and deploying stateful AI agents. The API is publicly available and supports sync, async, and streaming interaction modes across both cloud and self-hosted deployments. Auto-generated client SDKs are provided for Python and TypeScript, built from a unified OpenAPI specification using Fern. The API enables developers to manage agents, memory blocks, tools, conversations, and file operations. Documentation is available at docs.letta.com with endpoints for agent creation, memory management, tool integration, and conversational memory. Developer documentation is at docs.letta.com.
What industry is Letta in?
Letta's product category is AI Agent Development Platform. Its primary akta.pro industry code is HDAEANAJ, AI Application Enablement Platforms (Copilot/Agent Frameworks, SDKs), with a secondary code of BPAMAOAH, API Documentation & Developer Portals. Its NAICS code is 5132 and its SIC code is 7371.