Letta / MemGPT
Originated as MemGPT (UC Berkeley), where the model pages its own memory between context, recall and archival tiers. The product now centres on Letta Code, a stateful agent harness whose agents rewrite their own memory, skills and prompts over time.
Active · Verified 2026-10-29 · fresh as of 2026-10-29 · 75 cited facts, 4 with a third-party source · 2 recorded changes · Suggest a correction
At a glance
- Type
- Tiered: core / recall / archival vendor fix
- Tier
- T1
- Section
- Dedicated memory layers
- Latest release
- Letta Code v0.34.30 (npm @letta-ai/letta-code, 2026-10-29); PyPI letta v0.34.12 (2026-10-12); last letta-ai/letta server tag v0.16.8 (May 2026) registry fix
- License
- Apache-2.0 repo fix
- GitHub
- letta-ai/letta-code — ~3.7k stars, Apache-2.0 (current source: agent harness, terminal UI, App Server); letta-ai/letta (~25.3k stars, not archived) is now a README pointing to letta-code, with the retired Letta V1 API server on its archive branch repo fix
- Pricing
- Free $0 (3 agents); Pro $20/mo; Developer pay-as-you-go ($0.10/active agent/mo); Teams; Enterprise custom vendor fix
- Funding
- $10M total $70M val Seed · 2024-09 third-party fix
- Freshness
- 2026-10-29 fix
- Layer
- agent-infra repo fix
- SDK languages
- python, rest, typescript vendor fix
- Protocols
- mcp-client, mcp-server vendor fix
- Developer activity
- OpenRank 0.7 (12-mo avg 5.04) · 11 participants — 2026-09 third-party fix
Memory taxonomy
- storage
- kv
- retrieval
- similarity
- persistence
- long-term
- update
- agent-controlled
- unit
- episode
- governance
- inspectable
- conflict
- llm-arbitrate
When to use
Optimised for: LLM-as-OS paradigm; agent autonomy in managing its own memory
Anti-fit: no anti-fit explicitly stated
Pros & cons
Pros
Most architecturally principled — explicit OS-style tiering makes capacity behavior predictable; UC Berkeley origin + active v1 maturity.
Cons
Tier configuration adds operational complexity vs flat vector memory; latency is sensitive to paging decisions on long histories.
Claims & capabilities
$10M seed at $70M (Sep 2024). Reports ~83.2% on LongMemEval; ~74.0% on LoCoMo with gpt-4o-mini using files-as-memory.
Technical surface
- API surface
- REST, SDK: Python, TS
- Backend storage
- Postgres + pgvector
- Deployment
- Both
- Embedding model
- multiple supported
- Multi-tenancy
- Logical namespace per org/project/agent; self-hosted Docker option for hard tenant isolation
- MCP
- native (first-party) — Letta serves & consumes MCP
- A2A
- not documented publicly
- OpenTelemetry
- not documented publicly
Compare Letta / MemGPT with…
Alternatives to Letta / MemGPT → live options ranked by shared design, licence, hosting and MCP support
Similar systems
Other dedicated memory layers in the catalog, ranked by inbound references.
- Mem0 T1
Universal memory layer for AI agents. Three concurrent stores (vector + graph + KV); LLM-extracted facts; concurrent retrieval via ThreadPoolExecutor.
- Zep & Graphiti T1
Bi-temporal knowledge graph (event time + ingestion time). Strong on chronological reasoning and contradiction tracking. Graphiti is the open-source core.
- Cognee T1
"Extract–Cognify–Load" pipeline that turns raw input into a typed, queryable knowledge graph for agent recall.
- Hindsight (Vectorize) T1
Standalone memory service from Vectorize. Open source. Biomimetic four-network design (World, Bank, Observation, Opinion). Ships an MCP memory server.
- Memori T2
Structured, persistent memory layer for AI agents. Open source.
- Memvid T2
Single-file memory layer (one .mv2 file). No DB, no server. Append-only sequence of immutable Smart Frames with timestamps + checksums. Native Rust core (rewritten from Python).
Sources (74 cited facts, 4 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Tiered: core / recall / archival vendor
- What it does & what's distinct Originated as MemGPT (UC Berkeley), where the model pages its own memory between context… repo
- Claims / benefits $10M seed at $70M (Sep 2024). Reports ~83.2% on LongMemEval; ~74.0% on LoCoMo with gpt-4… vendor
- Modalities text vendor
- Created 2023-10 vendor
- Latest release Letta Code v0.34.30 (npm @letta-ai/letta-code, 2026-10-29); PyPI letta v0.34.12 (2026-10… registry
- License Apache-2.0 repo
- GitHub letta-ai/letta-code — ~3.7k stars, Apache-2.0 (current source: agent harness, terminal U… repo
- Funding $10M total $70M val Seed · 2024-09 third-party
- Customers / scale $1M ARR 11-50 ppl No named enterprise customers disclosed third-party
- Pricing Free $0 (3 agents); Pro $20/mo; Developer pay-as-you-go ($0.10/active agent/mo); Teams; … vendor
- Deployment Both vendor
- HQ US vendor
- Founders / pedigree UC Berkeley Sky Computing Lab spin-off; co-founders Charles Packer and Sarah Wooders (bo… third-party
- Memory volume / scale no scale claim vendor
- Performance 74.0% LoCoMo 83.2% LMES vendor
- API surface REST, SDK: Python, TS vendor
- Backend storage Postgres + pgvector vendor
- Multi-tenancy Logical namespace per org/project/agent; self-hosted Docker option for hard tenant isola… vendor
- Encryption AES-256 at rest, TLS in transit (Letta Cloud on managed cloud) vendor
- SSO / RBAC SSO + RBAC vendor
- Embedding model multiple supported vendor
- Consistency strong vendor
- Versioning snapshots vendor
- Tombstoning hard-delete supported vendor
- Schema evolution schema with migration vendor
- Namespace primitives per-user vendor
- Contradiction handling LLM resolves and overwrites vendor
- Forgetting policy consolidation/summarization vendor
- MCP support native (first-party) — Letta serves & consumes MCP vendor
- Import / export JSON agent file (.af) export/import vendor
- Integration count 1 repo
- Orchestration single-agent (per agent: core, archival, recall memory blocks) vendor
- Programmatic control full CRUD + SDK (Python/TS REST) vendor
- Vendor benchmarks LMES 83.2 (vendor); LoCoMo 74.0 with gpt-4o-mini (vendor) vendor
- Pricing specifics per-message + per-agent tier flat-rate: Letta Cloud Free; Pro $20/mo; Enterprise custom;… vendor
- Validated verticals consumer / individual, research, agent developers (horizontal) vendor
- Time to running <1hr (Docker compose, OSS) or hosted Letta Cloud vendor
- Anti-fit no anti-fit explicitly stated vendor
- Optimised for LLM-as-OS paradigm; agent autonomy in managing its own memory vendor
- Adjacent infrastructure BYO LLM (OpenAI/Anthropic/local); bundled storage (Postgres/SQLite) vendor
- Pros Most architecturally principled — explicit OS-style tiering makes capacity behavior pred… vendor
- Cons Tier configuration adds operational complexity vs flat vector memory; latency is sensiti… vendor
- Project & sources GitHub letta-ai/letta Letta docs v1 agent blog agent memory blog benchmarking vendor
- obs-opentelemetry yes vendor
- obs-custom First-party OpenTelemetry tracing in Letta Server vendor
- cost-token-budget yes vendor
- cost-prompt-caching no vendor
- cost-semantic-caching no vendor
- cost-batching no vendor
- cost-model-routing yes vendor
- cost-streaming-only no vendor
- cost-observability-cost-attribution no vendor
- eval-langsmith-evals no vendor
- eval-braintrust no vendor
- eval-weights-and-biases-agent no vendor
- eval-helicone-evals no vendor
- eval-custom-test-harness yes vendor
- eval-human-loop yes vendor
- eval-production-traffic-replay no vendor
- commit-trajectory [{"ym":"2025-11","cum":6578},{"ym":"2025-12","cum":6733},{"ym":"2026-01","cum":7043},{"y… repo
- download-trajectory [{"ym":"2025-11","monthly":9282,"cum":9282},{"ym":"2025-12","monthly":31216,"cum":40498}… registry
- cost-input-usd-per-mtok 0 vendor
- cost-output-usd-per-mtok 0 vendor
- cost-tier free vendor
- cost-pricing-model free-tier vendor
- cost-last-verified 2026-06-29 vendor
- architecture-deployment-locus hybrid repo
- licensing-bucket open-core vendor
- positioning-platform-vs-product both repo
- taxonomy-three-layer-rollup agent-infra repo
- developer-activity-opendigger {"activity":20.5,"as_of":"2026-09","openrank_12mo":5.04,"openrank_current":0.7,"particip… third-party
- sdk-languages python,rest,typescript vendor
- protocol-support mcp-client,mcp-server vendor
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)
- 2026-10-29
- Latest release “Letta Code (PyPI letta) v0.34.12 (2026-10-12); last letta-ai/letta server tag v0.16.8 (May 2026)” → “Letta Code v0.34.30 (npm @letta-ai/letta-code, 2026-10-29); PyPI letta v0.34.12 (2026-10-12); last letta-ai/letta server tag v0.16.8 (May 2026)” source
- 2026-10-22
- Latest release “0.16.7 2026-03-31” → “Letta Code (PyPI letta) v0.34.12 (2026-10-12); last letta-ai/letta server tag v0.16.8 (May 2026)” source
- Pricing “Free + paid” → “Free $0 (3 agents); Pro $20/mo; Developer pay-as-you-go ($0.10/active agent/mo); Teams; Enterprise custom” source
Related systems
References (11)
- Anthropic Claude (foundation models) integrates with · depends on at runtime — adjacent-infrastructure cell: BYO LLM (OpenAI/Anthropic/local); bundled storage (Postgres/SQLite)
- DeepSeek R1 / V3 family integrates with — Letta models docs list DeepSeek in the supported model provider table. (verified 2026-10-29)
- Fireworks AI integrates with — Letta models docs list Fireworks in the supported model provider table. (verified 2026-10-29)
- Google Gemini 3 family integrates with — Letta models docs list Google Gemini in the supported model provider table. (verified 2026-10-29)
- Mistral Large 2 / Mixtral family integrates with — Letta models docs list Mistral in the supported model provider table. (verified 2026-10-29)
- Model Context Protocol (MCP spec) integrates with — Letta Agent SDK docs cover connecting sessions to stdio, HTTP and SSE MCP servers. (verified 2026-10-29)
- OpenAI GPT family (GPT-5 / GPT-4o / o3 / o4) integrates with · depends on at runtime — adjacent-infrastructure cell: BYO LLM (OpenAI/Anthropic/local); bundled storage (Postgres/SQLite)
- OpenRouter integrates with — Letta models docs list OpenRouter in the supported model provider table. (verified 2026-10-29)
- pgvector depends on at runtime — backend-storage cell: Postgres + pgvector
- Together AI integrates with — Letta models docs list Together AI in the supported model provider table. (verified 2026-10-29)
- xAI Grok 4 integrates with — Letta models docs list xAI in the supported model provider table. (verified 2026-10-29)
Referenced by (5)
- e2b integrates with — E2B docs have a Letta / MemGPT integration guide. (verified 2026-10-29)
- LiveKit Agents integrates with — LiveKit Agents docs have a Letta / MemGPT model plugin page. (verified 2026-10-29)
- Milvus integrates with — Milvus's integrations overview lists a Milvus + Letta / MemGPT integration tutorial. (verified 2026-10-29)
- Pydantic Logfire integrates with — Pydantic Logfire integrations docs have a Letta page. (verified 2026-10-29)
- Vercel AI SDK integrates with — AI SDK docs have a Letta community provider page. (verified 2026-10-29)