DAKERASelf-hosted. Free to start.
Dakera is in early access. Run the full memory server on your own infrastructure today — free, with no limits. We're shaping our enterprise offering based on what teams actually need.
Self-Hosted
Everything you need to add persistent memory to your AI agents. Self-hosted, full control, production-ready.
Enterprise
We're building Dakera's enterprise offering in the open. Tell us what your team needs and help shape what we build.
Everything included, no paywalls
There's no feature-gated free tier. You get the full product.
Full Memory Engine
Hybrid retrieval combining BM25, vector search, and knowledge graph — all in a single Rust binary.
Built-in Embeddings
ONNX-based embedding model included. No external API calls, no per-token fees.
4 SDK Languages
Python, JavaScript, Rust, and Go. All MIT-licensed, all production-ready.
MCP Server
Works with Claude, Cursor, Windsurf, and any MCP-compatible client.
CLI Tools
Manage agents, query memories, and debug from the command line.
Your Infrastructure
Self-hosted. Your data never leaves your servers. No vendor lock-in.
Frequently asked questions
Is Dakera really free?
Yes. The memory engine is proprietary but free to self-host with no usage fees or time limits. SDKs, CLI, and MCP server are MIT-licensed open source.
Will there be a paid tier?
We're exploring enterprise offerings like dedicated support, deployment assistance, and custom feature development. The self-hosted product will remain free.
Do I need to pay for embeddings?
No. Dakera includes a built-in embedding model. You never need an external embedding API or pay for embedding tokens.
Where does my data live?
On your servers. Dakera is self-hosted — your data never leaves your infrastructure. Storage uses local files on your server.