For builders

Start with one repository, one docs folder, or one internal agent workflow.

Keep your database and agent framework. Use Cortrix to evaluate the paths that usually become glue code: ingestion scripts, memory stores, retrieval feedback, audit context, and workflow state.

Retrieval

Stop rebuilding the same pipeline

Use Cortrix to evaluate ingestion, semantic search, reranking, and agent-facing query paths in one open-source storage layer without treating it as a full replacement for every existing system.

Memory

Keep useful context available

Agent workflows need more than one-off retrieval calls. Cortrix gives builders a path to persist, search, inspect, and invalidate interaction memory.

Audit

Trace decisions back to sources

For debugging and review, Cortrix focuses on source-level provenance: which documents, chunks, and turns contributed to an answer.

Builder paths

The first public routes are organized around the practical questions builders ask before adopting a new agent data layer.

New project

Run Cortrix locally, query the demo namespace, then connect through the HTTP API, Python SDK, MCP server, or framework toolkit.

Existing stack

Use /migration/ to evaluate where Cortrix can reduce glue code without forcing a replacement of your operational database or agent framework.

Evaluation

Use /benchmark/ for retrieval-quality results, runner methodology, artifact format, and claim boundaries.