LIFEHUBBER
Theme

AI Resources

ktx

ktx is a local context layer for data agents, built to help AI coding and agent tools query SQL warehouses with reusable metric definitions, warehouse context, wiki knowledge, and MCP or CLI access.

The repository frames ktx around approved metric definitions, joinable columns, semantic-layer entities, business wiki content, data-stack scanning, local project files, agent setup, and a local MCP daemon for agent clients. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.

What it is

Context infrastructure for data agents

ktx is positioned as a layer that helps agents find and reuse warehouse context instead of starting from scratch on every data question, with local project files for semantic sources, wiki notes, and generated context.

Why it stands out

Warehouse context plus agent-facing tools

The README highlights SQL warehouse support, dbt and semantic-layer ingestion, wiki or Notion-style business context, contradiction flags, CLI commands, and MCP tools for agent execution.

Availability

GitHub repo, npm package, and docs

The public materials include a GitHub repository, npm install path, quickstart commands, CLI reference, agent setup docs, MCP startup notes, project-layout guidance, and development instructions.

Why it matters

Why readers may notice it

Data agents often need more than raw database access. They need the surrounding business meaning: which metrics are approved, which joins make sense, where definitions live, and what context should be reused instead of guessed again.

Reporting note

What appears notable

The README is useful for checking warehouse scanning, approved metric definitions, joinable-column detection, wiki ingestion, CLI search commands, MCP server support, and the local project layout with `.ktx/` kept out of git.

Before using

What readers may want to review

Which warehouse, semantic-layer, wiki, and business-context sources the project will be allowed to read.

Which LLM or embedding provider is configured, since the README says provider-bound data depends on the user's chosen setup.

Telemetry, project files, local secrets, and company data rules before connecting internal sources.

Whether the team already has approved metric definitions or data-governance rules that ktx should follow rather than replace.

Reader fit

Who may find it relevant

Data teams exploring how AI agents can query warehouses without inventing metric logic each time.

Builders comparing MCP tools, CLI context search, semantic-layer ingestion, and wiki-backed agent context.

Teams using agents such as Codex, Claude Code, Cursor, or OpenCode around analytics projects.

Less relevant for readers who do not use a SQL warehouse or only need a one-off notebook query.

Editorial note

Why it is included here

ktx gives readers a practical comparison point for a practical context layer for data agents.

Source links

Original materials

Reader note

Before relying on this entry

LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.

Related in LifeHubber

Keep the thread going

Follow the next layer with AI Resources for AI projects worth inspecting at the source, AI Guides for decision habits for messy AI choices, AI Access for free and low-cost ways to compare AI model access, AI Ballot for a clearer view of what readers are leaning toward, and AI Radar for AI stories that deserve a second look.