Skip to main content

Get Started with Cognee

Quickstart your journey with Cognee. Install, configure, and run your first example to build AI memory applications.
Diagram showing raw documents becoming chunks, extracted entities, derived concepts, induced ontologies, and searchable memory through remember, improve, and recall. Diagram showing raw documents becoming chunks, extracted entities, derived concepts, induced ontologies, and searchable memory through remember, improve, and recall.

​
Choose the best product for you

All four build the same kind of memory — a knowledge graph you fill with remember and query with recall — so the choice comes down to where that memory runs and what you build in.

Cognee

Python, self-hostedThe open-source engine, and the place to start if you are not sure: the Python library, the CLI, the MCP server, the HTTP API, and direct integrations. Runs on your own machine with embedded databases, or on your own infrastructure.

Cognee Cloud

Hosted, nothing to installThe same engine, managed: a web UI, preconfigured storage, pipeline execution, and dataset-level permissions for a team. Reach it from the Python SDK, the HTTP API, or MCP — or run the same UI locally for free, with no account.

Rust SDK

On-device, experimentalcognee-rs builds memory pipelines for phones, watches, and embedded targets, driven by a cognee-cli binary. You build it from source with Cargo — there is no hosted or pip install path.

TypeScript SDK

Node.js@cognee/cognee-ts brings the same memory API to Node.js through native bindings over cognee-rs. Install it from npm and point it at local memory or at your Cognee Cloud tenant.