Cognee
Open-source memory for AI agents, accessible through SDK, HTTP, and MCP with hosted or private deployment options.
About Cognee
Cognee is an open-source memory platform for AI agents that combines vector search, knowledge graphs, and relational storage. Its memory API centers on four operations: remember, recall, improve, and forget. Teams can use the Python SDK, HTTP API, or MCP to give agents context that persists across conversations and sessions. Cognee can run locally, in Docker, on premises, in a private cloud, or as a managed cloud service. Teams can keep data within their own environment, then move to managed infrastructure if their needs change. The hosted Free plan includes one workspace, 1 million processed tokens, unlimited users, and unlimited API calls. Standard charges $2.50 per 1 million tokens and $5 per additional workspace each month. Enterprise adds dedicated support, bring-your-own-cloud deployment, and a service-level agreement (SLA). Self-hosting is free, but teams must operate the infrastructure and account for model, database, and maintenance costs themselves.
Best for
Best use cases
Choose Free for a first hosted agent within one workspace and 1 million tokens
Choose Standard when token usage or multiple workspaces grows
Discuss Enterprise when dedicated support, a support SLA, or controlled cloud deployment is required
Compare before choosing
Hosted pricing is tied to processed tokens rather than API-call count
Self-hosting avoids a Cognee license fee but shifts infrastructure and operations to your team
Standard adds paid workspaces and data-source integrations that the Free plan does not include
Features
Persistent agent memory
Store context that agents can recall across conversations and sessions through remember and recall operations.
GraphRAG and connected retrieval
Combine vector similarity with graph relationships and relational data when agents need connected context instead of isolated chunks.
Private AI memory deployments
Run the memory engine locally, in Docker, on premises, or in a private cloud when data boundaries rule out a shared managed service.
Pros
- A free hosted plan includes 1 million tokens, one workspace, unlimited users, and unlimited API calls.
- SDK, HTTP, and MCP access lets teams connect memory to different agent workflows without adopting one client interface.
- Local, on-premises, private-cloud, and bring-your-own-cloud options give organizations control over where memory data runs.
Cons
- Standard charges by processed token volume, so costs rise with ingestion and memory workload.
- Each additional Standard workspace costs $5 per month, which adds a fixed cost for separated teams or projects.
- Self-hosting removes the Cognee license fee but leaves infrastructure, model services, monitoring, and maintenance with your team.
Pricing Plans
Free
- 1 million tokens included
- 1 workspace
- Unlimited users
- Unlimited API calls
- Agentic integrations including Claude Code, Codex, and MCP
Standard
- Everything in Free
- Unlimited workspaces at $5 each per month
- Slack, Notion, and Google Drive data source integrations
- In-app support
Enterprise
- Everything in Standard
- Dedicated Slack channel
- Dedicated support engineer
- Bring-your-own-cloud support
- Support SLA
* Pricing information is for reference only. Please refer to the latest announcement on Cognee's official website.