Cognee

Cognee is an open-source AI memory engine built by topoteretes, headquartered in Berlin, that transforms raw data into structured knowledge graphs for AI agents. While most memory systems store text chunks in a vector database, Cognee runs an ECL pipeline (Extract, Cognify, Load) that extracts entities, maps relationships, and builds a queryable knowledge graph with embeddings: giving agents temporal awareness, entity relationships, and feedback loops that pure vector retrieval cannot provide. The core operations are four: remember (store to graph), recall (query with auto-routing), forget (delete), and improve (refine through feedback). When an agent rates a response, that feedback updates edge weights in the graph: the memory gets sharper with use rather than staying static. Cognee unifies three storage layers (relational, vector, and graph) into a single engine using SQLite, LanceDB, and Kuzu locally, with managed cloud options for production scale. $7.5M seed raised in February 2026 from investors including former OpenAI and Facebook AI Research builders. Pipeline volume grew from 2,000 runs to over 1 million runs in 2025. 500x growth. 17.3k+ GitHub stars as of May 2026, 1.8k forks. In production at 70+ companies including Bayer for scientific research workflows. Graduated from the GitHub Secure Open Source Program. Native integrations with Claude Agent SDK, OpenAI Agents SDK, LangGraph, Google ADK, n8n, Amazon Neptune, Neo4j, and more. Official MCP integration available via docs. Available as a Claude Code plugin. Pricing: Free tier (open-source self-hosted or hosted cloud free tier) covers memory workflows, auto-generated knowledge structures, and 28+ data sources with community support. Developer plan at $35/month includes 1,000 documents (approx 1 GB), 1 user, full API endpoints, automated scaling, parallel processing, and 10,000 API calls per month with top-up packs available (1,000 docs for $35, 3,000 docs for $100, 15,000 docs for $750). Cloud Team plan at $200/month includes 2,500 documents (approx 2 GB), up to 10 users, multi-tenant architecture, and everything in Developer. Enterprise is custom-priced. Graph features are available at every tier: unlike Mem0, which gates graph memory behind the $249/month Pro tier, Cognee provides full graph capabilities from the free tier. Who Cognee is not for: Python-only SDK means TypeScript, Go, or other language-based agent stacks cannot use Cognee natively without custom wrappers, whereas Mem0 supports both Python and JavaScript. Teams that need a turn-key hosted memory solution with minimal setup may find Mem0 Cloud simpler to onboard. Teams requiring published temporal reasoning benchmarks for procurement comparison: Cognee has not published a LongMemEval score comparable to Mem0 (49%) or Hindsight (91.4%), making objective capability comparison harder. Alternatives: Mem0 for cross-language support and a larger community; Zep for structured user memory with session history; custom vector database solutions (Pinecone, Weaviate) for teams that do not need graph-based relationship mapping. Current state Q2 2026: Cognee has surpassed 5 million SDK runs per month and is in production at 70+ organizations including Bayer for scientific research workflows. The project is part of the Berkeley XCelerator program. Version 1.1.2 released May 2026. GitHub stars reached 17.6k with 1.9k forks. The official MCP server provides 14 specialized tools compatible with Claude Desktop, Cursor, Continue, Cline, and Roo Code, supporting both standalone mode for individual developers and API mode for teams sharing a single knowledge graph across multiple AI clients.