Mem0
Mem0Mem0 adds long-term memory to AI agents and assistants. It extracts facts from conversations with a language model, stores them with embeddings, and returns the relevant ones by user, agent or session before the next model call. Developers run it as a Python or TypeScript library, a self-hosted server, or the hosted Mem0 Platform, all built by the company of the same name.
Openness
4 high confidence- license
- Apache-2.0 (mem0 Python library, mem0-ts TypeScript SDK and self-hosted server)
- source
- public
- core features withheld
- yes — graph memory, memory decay, temporal reasoning and Dream consolidation exist only in the hosted Mem0 Platform, and the open SDK raises "not supported by the OSS Memory SDK" for decay, timestamp and reference_date; the 1,844-path tree at fff8a64 has no ee/ or enterprise/ path, because the withheld features are not shipped at all
- commercial
- Mem0-Platform (hosted service with graph memory, decay, temporal reasoning, Dream, webhooks and org and project scoping)
Mem0's memory library, TypeScript SDK and self-hosted server are published under Apache-2.0 and run without a Mem0 account. Several memory features exist only in the hosted Mem0 Platform, though: the entity graph, decay of stale memories, temporal reasoning and Dream background consolidation, and the open SDK returns an error when asked for decay or temporal parameters. A paid service holding back part of the memory engine makes this open core.
- https://docs.mem0.ai/platform/platform-vs-oss recorded 2026-11-20
The comparison page lists Graph Memory as "Removed" from open source, Memory Decay and Temporal Reasoning as "Not supported", and Dream (background consolidation), webhooks, memory export and batch operations as "Not available" outside the Platform.
- https://mem0.ai/pricing recorded 2026-11-20
Pricing offers a Free plan and a Custom plan; the Custom plan adds "Graph memory (entity linking)", "User profiles", "Dream (memory consolidation)" and "On-prem deployment".
- https://raw.githubusercontent.com/mem0ai/mem0/main/LICENSE recorded 2026-11-20
The root LICENSE is the Apache License 2.0 text with "Copyright [2023] [Taranjeet Singh]".
- https://raw.githubusercontent.com/mem0ai/mem0/main/mem0/memory/notices.py recorded 2026-11-20
The open SDK's notices module defines "The decay parameter is not supported by the OSS Memory SDK." and the same message for the timestamp and reference_date parameters.
- https://raw.githubusercontent.com/mem0ai/mem0/main/pyproject.toml recorded 2026-11-20
The mem0ai package manifest declares license = "Apache-2.0" and builds from the repository's mem0 directory.
- https://raw.githubusercontent.com/mem0ai/mem0/main/README.md recorded 2026-11-20
The README offers a library ("pip install mem0ai"), a self-hosted server ("docker compose up") and the Cloud Platform, marks Advanced Features as "Teasers" on the self-hosted server and "All included" on the Platform, and says benchmark scores "reflect Mem0's managed platform, which includes proprietary optimizations not available in the open-source SDK".
Adoption
4 medium confidenceAdoption is counted as monthly downloads of the mem0ai package on PyPI and on npm. Each package carries both the self-hosted library and the client for the hosted Platform, so the count cannot separate open-source use from Platform use.
- https://api.npmjs.org/downloads/point/last-month/mem0ai recorded 2026-11-20
1,041,937 downloads of mem0ai between 2026-10-20 and 2026-11-18.
- https://pypistats.org/api/packages/mem0ai/recent recorded 2026-11-20
2,544,949 downloads of mem0ai in the last month.
- https://registry.npmjs.org/mem0ai/latest recorded 2026-11-20
The npm package's repository is git+https://github.com/mem0ai/mem0.git, directory mem0-ts, license Apache-2.0.
Capability
2 medium confidenceMem0's open library extracts facts from conversations and serves them back by user, agent or session, but it only ever adds memories: it does not merge or correct what is stored, keeps no graph, and forgets only on dates the caller sets. Consolidation, decay and the entity graph are features of the hosted Mem0 Platform, not of the code Mem0 publishes.
- https://docs.mem0.ai/core-concepts/how-it-works recorded 2026-11-20
"An LLM extracts preferences, decisions, plans, and other details"; searches are scoped by user_id, agent_id or run_id; "The automatic extraction path is additive", and OSS retrieval "has no graph memory".
- https://docs.mem0.ai/platform/features/dream recorded 2026-11-20
Dream "synthesizes higher-order patterns, supersedes outdated facts, and merges duplicates"; Supersede and Merge are "Always on, all plans" of the Platform.
- https://docs.mem0.ai/platform/features/graph-memory recorded 2026-11-20
"Mem0 Platform automatically organizes your memories into a graph": entities become nodes and memories that share an entity are connected, boosting search ranking.
- https://docs.mem0.ai/platform/features/memory-expiration recorded 2026-11-20
"Set an expiration date on a Mem0 memory and it stops surfacing in search once that date passes." "Works on Platform and Open Source."
- https://docs.mem0.ai/platform/platform-vs-oss recorded 2026-11-20
Memory Decay "reinforces recently-used memories and gently dampens stale ones at search time" on the Platform and is not supported in open source.
Verified 2026-11-20, except some scores, still awaiting re-confirmation