Skip to content
Letta Docs
Search
Ctrl
K
Auto
Light
Dark
Sign up
Guides
Tutorials
API Reference
Letta Code
Guides
Tutorials
API Reference
Letta Code
Get started
Overview
Quickstart
Core concepts
Letta SDK Reference for AI Agents
Agents
Building stateful agents
Context hierarchy
Image inputs
Structured output
Multi-agent systems
Multi-agent patterns
The supervisor-worker pattern
Parallel execution with shared memory
The producer-reviewer pattern
The round-robin pattern
AgentFile (.af)
Memory
Memory overview
Memory blocks
Shared memory
Archival memory
Archival memory
Searching & querying
Best practices
Exporting archival
Context engineering
Compaction settings
Tools
Connecting agents to tools
Web search
Code interpreter
Filesystem
Tool types
Base tools
Custom tools
Utilities
Tool execution
Tool execution modes
Sandbox execution
Client-side tool execution
MCP remote execution
Built-in tool execution
Local tool execution
Model Context Protocol
What is MCP?
Connect Letta to MCP
Remote MCP servers
Local MCP servers
Advanced usage
Multi-agent tools
Creating tool rules
Using tool variables
Parallel tool calling
Deploy and scale
Message types
Streaming
Long-running executions
Scheduling
Human-in-the-loop
User identities
Role-based access control
Client-side access tokens
Integrations
Claude Code memory proxy
Development tools
Letta ADE
ADE overview
Connecting to the ADE
Accessing the web ADE
Installing Letta Desktop
Agent simulator
Context window viewer
Core memory
Archival memory
Data sources
Tools
Agent settings
Community tools
lettactl
Letta Snippets
Testing & evals
Letta Evals
Getting started
Core concepts
Core concepts
Suites
Datasets
Targets
Graders
Extractors
Gates
Configuration
Suite YAML reference
Graders
Tool Graders
Rubric Graders
Multi-metric evaluation
Extractors
Built-in extractors
Custom extractors
Advanced
Custom Graders
Multi-turn conversations
CLI reference
CLI commands
Results & metrics
Understanding results
Troubleshooting
Troubleshooting
Templates & versioning
Intro to templates
Template versioning
Memory variables
Self-hosting
Overview
Local tool execution
Collecting traces
Model providers
OpenAI
Anthropic
Google AI (Gemini)
LM Studio
More providers
OpenAI-compatible
DeepSeek
AWS Bedrock
Groq
xAI (Grok)
Together
Google Vertex AI
Azure OpenAI
Ollama
vLLM
Advanced configuration
Database configuration
Performance tuning
Inspecting DB
Installing from source
Remote hosting
Deploy a server
Deploy on Railway
Experimental
Sleep-time agents
Conversations
Voice agents
Voice agents
LiveKit integration
Vapi integration
Additional resources
Letta FAQs
Letta Desktop FAQ
Letta ADE FAQ
Research background
Research background
MemGPT
Legacy & migration
Legacy agent types
Architecture migrations
MemGPT agents
Workflows
Low-latency agents
ReAct agents
Heartbeats
Agent types naming
Auto
Light
Dark
Utool uexecution
Configure tool execution environments and security settings for self-hosted deployments.
