Kogcat Mcp
Ground agent answers in your local knowledge base and surface counter-examples before you commit to a spec or ship a risky change.
Overview
KogCat MCP is a MCP server for the Validate phase that queries your local knowledge base to expose counter-examples and blind spots while the agent plans or reviews work.
What is this MCP server?
- Local-first queries over your own knowledge base—no cloud corpus required for the judgment pass.
- Surfaces counter-examples and blind spots instead of only affirming the model’s first answer.
- Stdio MCP package (kogcat-mcp on PyPI) for Claude Code, Cursor, and other stdio-capable agents.
- Version 0.46.0 server schema with GitHub source at KogCat/cc-kogcat.
- Fits validate-and-ship loops: challenge plans, then re-run before merge or launch.
- Server version 0.46.0
- PyPI package identifier kogcat-mcp with stdio transport
- Repository: github.com/KogCat/cc-kogcat
What problem does it solve?
Agents sound confident on ideas your own notes already contradict, and you only notice after you have already built or shipped.
Who is it for?
Solo builders who keep specs, postmortems, or research in a local KB and want the agent to argue with its first draft before coding.
Skip if: Teams that need live web fact-checking or have no curated local corpus to query.
What do I get? / Deliverables
Your agent routinely challenges proposals with counter-examples drawn from knowledge you control, so scope and review decisions stay aligned with what you already documented.
- Counter-example snippets tied to your KB
- Blind-spot callouts during agent turns
- Repeatable judgment passes across validate and review
Recommended MCP Servers
Journey fit
Blind-spot checks belong earliest in validate when scope is still flexible, but the same judgment layer helps during build decisions and pre-ship review. Scope is where unstated assumptions hurt solo builders most; KogCat tags gaps with counter-examples from material you already trust.
How it compares
MCP judgment layer over your KB—not a generic web search or a single prompt skill.
Common Questions / FAQ
Who is com.kogcat/kogcat-mcp for?
Indie developers and small teams using Claude Code or Cursor who want agents to sanity-check plans against their own documented knowledge.
When should I use com.kogcat/kogcat-mcp?
Use it when scoping a feature, evaluating an AI-generated architecture, or running a pre-ship review where blind spots are costly.
How do I add com.kogcat/kogcat-mcp to my agent?
Install the PyPI package kogcat-mcp (v0.46.0), add a stdio MCP server entry pointing at that binary, restart the agent, and index or point tools at your local knowledge base per the cc-kogcat repo.