Over Reach Detector
Catch when an AI agent’s diffs wander beyond the task you declared, before you merge or ship unintended changes.
Overview
Over-Reach Detector is a Ship-phase MCP server that detects when AI-generated code changes exceed the scope of a declared task.
What is this MCP server?
- Over-Reach Detector MCP server v0.0.6 distributed as PyPI package over-reach-detector with uvx runtime hint.
- stdio MCP transport for hooking into Claude Code, Cursor, or other hosts that support Python MCP servers.
- Purpose-built to flag AI code changes that exceed declared task scope.
- Lightweight governance layer you run alongside agent editing sessions—not a full static analyzer.
- GitHub source at choreoatlas/over_reach_detector for audit and local development.
- MCP server version 0.0.6 on PyPI identifier over-reach-detector
- stdio transport with uvx runtime hint
- Single published package entry in server.schema (no required env vars in manifest)
What problem does it solve?
Agents often touch files and behaviors you never authorized, and manual diff review does not scale when you ship fast solo.
Who is it for?
Solo builders running heavy agent refactors who want an MCP hook to sanity-check scope against the stated task.
Skip if: Teams needing deep SAST, license compliance, or security vulnerability scanning instead of task-boundary checks.
What do I get? / Deliverables
After you add the stdio server, your agent workflow can call detection tooling that surfaces over-scope edits before you accept or deploy them.
- Registered stdio MCP server from PyPI over-reach-detector 0.0.6
- Agent-accessible checks that highlight changes beyond declared task scope
- Repeatable scope review step in Ship before merge
Recommended MCP Servers
Journey fit
Scope discipline is a Ship concern—this server supports review and launch prep when agent-generated patches need a guardrail against creep. It fits code review subphase because it compares declared intent to actual change sets rather than authoring features.
How it compares
Scope-creep detector MCP, not a full code-review skill or security pentest suite.
Common Questions / FAQ
Who is Over Reach Detector for?
Indie developers and small teams using AI agents who need an automated signal when edits go past the task they defined.
When should I use Over Reach Detector?
Use it during Ship review whenever an agent produces a patch and you want to verify the diff matches declared scope before merge or release.
How do I add Over Reach Detector to my agent?
Register the PyPI package over-reach-detector (v0.0.6) as a stdio MCP server—typically via uvx—in your Claude Code or Cursor MCP configuration.