Letta
Open-source platform for stateful AI agents with persistent memory. Build, observe, and deploy agents via ADE, REST API, SDKs, and a desktop app. Connect tools via MCP; use any model; run self-hosted or on Letta Cloud.
At a Glance
Pricing
Get started with Letta at no cost with 50 premium requests / month and 500 standard requests / month.
Engagement
Available On
About Letta
Letta is a platform for building stateful AI agents with transparent, persistent memory (rooted in the MemGPT research). It ships a developer-first Agent Development Environment (ADE) for designing, observing, and editing agent state in real time, plus a REST API and Python/TypeScript SDKs to deploy agents as microservices. Agents are model-agnostic, support custom tools, and can connect to external tool libraries through Model Context Protocol (MCP).
You can self-host from the open-source repo or use Letta Cloud to create, version, and operate agents at scale with templates, monitoring, and observability. A cross-platform Letta Desktop app is available for local development. The platform supports BYO LLM API keys, filesystem-backed knowledge, multi-agent patterns with shared memory, and portability via an agent file format.
Demo Video
Community Discussions
Be the first to start a conversation about Letta
Share your experience with Letta, ask questions, or help others learn from your insights.
Pricing
Free Plan Available
Get started with Letta at no cost with 50 premium requests / month and 500 standard requests / month.
- 50 premium requests / month
- 500 standard requests / month
- 100 active agents
- 2 agent templates
- 1 GB storage
Pro
Professional plan with 500 premium requests / month and 5,000 standard requests / month for power users.
- 500 premium requests / month
- 5,000 standard requests / month
- 10,000 active agents
- 20 agent templates
- 10 GB storage
- Monitoring & observability (cloud)
Scale
Scale plan with 5,000 premium requests / month and 50,000 standard requests / month.
- 5,000 premium requests / month
- 50,000 standard requests / month
- 10,000,000 active agents
- 100 agent templates
- 100 GB storage
- Advanced ops & team workflows
Enterprise
Enterprise-grade solution with Custom quotas and private model deployments and SAML/OIDC SSO and dedicated support.
- Custom quotas and private model deployments
- SAML/OIDC SSO
- Role-based access control (RBAC)
- BYOC deployment options
- Dedicated support
Capabilities
Key Features
- Persistent agent memory with transparent context management
- Agent Development Environment (ADE) for live inspect/edit of state, tools, and prompts
- Agents exposed as REST endpoints with auth/identities
- Python and TypeScript SDKs
- Model-agnostic: use OpenAI, Anthropic, Gemini, local (Ollama/vLLM)
- MCP integration for external tools and data sources
- Custom tools with sandboxed execution (cloud) and local tools (self-hosted)
- Agent templates, versioning, and rollbacks
- Monitoring and observability for production agents
- Self-hosting via Docker/pip or fully-managed Letta Cloud
- Desktop app for Mac/Windows/Linux (local agents)
- Filesystem and sources for document/knowledge ingestion
- Multi-agent systems with shared or linked memory
- Sleep-time/background workflows and request controls (tool rules)
- Portability between open source and cloud via agent file