FAQ

cognee is an open-source memory platform for AI agents. It unifies vector search, knowledge graphs, and relational storage into one engine, turning your data into connected, queryable memory that agents can recall across sessions.

cognee ingests your data, builds a knowledge graph with automatically generated (or custom) data models, and stores it across vector, graph, and relational layers — so agents retrieve connected context, not just similar chunks.

Teams use cognee to build agent memory, GraphRAG pipelines, semantic search, document intelligence, customer support agents, and internal copilots that need persistent context.

A vector database retrieves similar chunks. cognee combines vectors with graph relationships, generated data models, and memory operations so agents can retrieve connected context and reason across sources.

cognee works with the tools you already build on — Claude Code, Cursor, Codex, LangGraph, n8n, CrewAI, the OpenAI Agents SDK, and Google ADK — plus an MCP server for any MCP-compatible client.

Yes. On Cognee Cloud, memory is self-improving — it learns from new data and interactions so retrieval gets more accurate as your agents are used.

Yes. Cognee Cloud is multi-tenant with isolated workspaces and per-session memory, so you can serve many users or customers from one deployment.

cognee is state of the art on BEAM, the agent memory benchmark — outperforming the previously reported state of the art on long-term, multi-session recall (79% vs. 73.4% at 100k tokens, 67% vs. 64.1% at 10M), without building a custom architecture for the benchmark.

Yes. cognee is built to keep memory costs flat as data grows — in our benchmarks it processes the same workloads at a fraction of a baseline LLM approach's cost.

cognee connects to the data sources your team already uses — Slack, Notion, Google Drive, GitHub, Confluence, Jira, Dropbox, Amazon S3, Salesforce, and more. Under the hood it works with graph and vector backends including Kuzu, NetworkX, Neo4j, FalkorDB, LanceDB, Qdrant, Milvus, Weaviate, pgvector, and Redis, so you can match your existing infrastructure.

Yes. cognee runs on Postgres with pgvector, so you can add agent memory to infrastructure you already operate, instead of standing up a new database.

Yes. The cognee SDK is open source on GitHub, and teams can also use hosted Cognee Cloud or enterprise deployment options.

Yes. cognee can run locally for development and can be self-hosted for teams that need infrastructure control, private deployments, or stricter data boundaries.

cognee is GDPR-compliant, encrypted at rest and in transit, and supports air-gapped and bring-your-own-cloud deployments. Enterprise plans add SLAs, a dedicated support engineer, and a private Slack channel.

Cognee Cloud runs on gpt-oss-120b, OpenAI's open-weight model. Self-hosting the open-source engine lets you bring your own model.

Cognee Cloud is usage-based: a free tier with 1M tokens and one workspace, then $2.50 per 1M tokens on the Standard plan (extra workspaces $5/month each). Enterprise adds dedicated support, SLAs, and bring-your-own-cloud. The open-source engine is free to self-host, forever.

The fastest start is Cognee Cloud — create a free account at platform.cognee.ai (1M tokens, no card required). Prefer your own infrastructure? Self-host the open-source engine for free. The docs cover implementation, and the blog and academy go deeper on AI memory and knowledge graphs.

Join our vibrant community on Discord to chat with the team and other builders, explore the community page, and see where we'll be next on the events page. For partnerships, custom deployments, or enterprise questions, book a demo with our team directly.