Cognee open-sources agent memory platform: ontology from Postgres, 91% on LoCoMo
nikola_mr64990 · x · 2026-11-20
Cognee, an open-source memory platform for AI agents, lets agents remember, organize, and share knowledge without graph databases or proprietary models.
- Uses open-weight models + Postgres, with retrofitted embeddings instead of graph databases
- Scores 91% accuracy on LoCoMo and 79% on BEAM benchmarks
- Connects GitHub, Slack, Linear, Notion and 10+ other sources, auto-building a company ontology so agents don't burn millions of tokens just reading files
- Ships a small-model runtime, Cognee Cloud, and editable auto-ontologies with shared memory across agents, users, and teams
Related event: Cognee Builds Agent Memory on Postgres Without Graph Databases(2 posts)→
More from coding & agent
- lumimax tested: page-level citations, catches contradicting RAG papers — dr_cintas · 2026-11-20
- Running ComfyUI fully via Claude in a container — no manual UI anymore — xWranglerWranglerx · 2026-11-20
- AI writes all code to make a real robotic arm play chess, open-sourced — FinanceYF5 · 2026-11-20
- trace2rules audits AGENTS.md against real sessions: 66 rules, only 9 measurable, cargo fmt ignored 100% — rkkautsar · 2026-11-20
- Amazon's CompoundingKernel Hits #1 on SOL-ExecBench with Self-Improving GPU Kernel Agents — AmazonScience · 2026-11-20
- Indie dev builds a personal bot with Fable 5.2 + T3 Code, linking X and Meta data via a shared knowledge base — yihui_indie · 2026-11-20