Letta
Platform for building stateful AI agents with advanced memory that can learn and self-improve over time.
Letta
Letta (formerly MemGPT) is an open-source platform for building stateful agents with persistent memory. Unlike stateless LLM calls, Letta agents remember past interactions, accumulate knowledge over time, and can self-edit their own memory — enabling agents that genuinely improve with use.
Key Features
- Persistent memory layers: Core memory (always in context), archival memory (vector-searched), and recall memory (conversation history)
- Self-editing memory: Agents can autonomously read and write their own memory banks via tool calls
- Agent server: REST API + Python/TypeScript SDKs for integrating into any application
- Multi-agent orchestration: Agents can spawn sub-agents and communicate via a shared message bus
- Model-agnostic: Works with any OpenAI-compatible provider (GPT, Claude, Llama, Mistral)
- Sleeper agents: Agents persist between sessions with full state restoration
Quick Start
pip install letta
letta server # starts the agent server on :8283
from letta import create_client
client = create_client()
agent = client.create_agent(name="my-assistant")
response = client.send_message(
agent_id=agent.id,
message="Remember that I prefer concise answers.",
role="user"
)
Add to ai-supply
npx ai-supply add letta-stateful-agents
Curated mirror of the open-source Letta (Apache-2.0). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/letta-stateful-agents/check). Click a policy:
Consume Letta programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/letta-stateful-agents
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/letta-stateful-agents/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add letta-stateful-agents
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/letta-stateful-agents/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "letta-stateful-agents" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.