Favicon of ktx

ktx

An open-source context layer gives data agents approved metrics and business knowledge through local files, MCP tools and read-only database access.

Screenshot of ktx website

ktx gives analytics agents shared metric definitions and business context before they answer questions or write SQL. It's for teams that need agents to reuse approved calculations rather than invent their own each time.

  • Approved metric queries: Agents can compile measures, joins, filters and dimensions into SQL, with join handling that helps prevent inflated or miscounted results.
  • Context from existing systems: It combines warehouse metadata, query history, BI definitions and company documents, removes duplicate knowledge and flags contradictions for human review.
  • Reviewable files: Business knowledge lives in Markdown and metric definitions in YAML, so teams can review changes and keep them in Git alongside source evidence and caveats.
  • Agent access: Claude Code, Codex and Cursor can search context and run read-only database queries through MCP or a CLI; ktx never writes to connected databases.

Database connections include PostgreSQL, Snowflake, BigQuery, DuckDB and MongoDB. Integrations include dbt, MetricFlow, Looker, Metabase, Notion and Google Drive.

ktx is open source under Apache 2.0 and keeps its context project locally. It uses your LLM API keys or existing Claude Code or Codex authentication; local files don't imply offline model processing. Usage telemetry has opt-out controls, and error reports can contain local file paths.

Similar to ktx