MCP SERVERGive it the prompt you wrote and the diff the agent produced, and it names every change you did not ask for — the new dependency, the new env var, the endpoint, the cron job.
LABSCO SUMMARY
The single audit call is the whole product: it compares what the prompt authorized against what the diff actually contains, so a login-form request that arrives with a payment dependency and a checkout route surfaces as HIGH findings with the evidence attached. Only the prompt-reading step involves a model at all, and it is skippable — run it keyless and the extraction is regex, which is what makes two runs over the same diff agree.
A scope auditor for AI-written code. One call takes the originating prompt and a diff and returns structured findings plus a scope_creep_score, worked out by set arithmetic rather than a second model's opinion: what the diff contains, minus what the prompt authorized.
Nothing mandatory. It starts as npx -y overreach --serve. A key for the model that reads your prompt is optional — ANTHROPIC_API_KEY or OPENAI_API_KEY if you want the scope extracted by a model, none at all for the offline regex mode, and none for a local Ollama. Anonymous usage pings carry counts and enums only, never code, prompts, file paths or diff content, and OVERREACH_TELEMETRY=0 or DO_NOT_TRACK=1 turns them off. The same package also runs as a CLI and a GitHub Action that exits non-zero when unexpected changes are found. MIT.
One command — npx overreach --serve
USE CASES
Pass the prompt and the diff and get the out-of-scope changes ranked, each with its file and the evidence that triggered it — new env vars, endpoints, cron jobs and listeners first.
Additions to package.json that no part of the prompt authorized come back as scope.dep findings, alongside any files touched outside what the prompt described.
The agent audits its diff against the instruction it was given and reads the score back; nothing is sent anywhere when it runs in keyless mode.
The blast-radius pass flags a schema change with no migration, a new route with no test, a package.json without its lockfile, and secrets written straight into source.
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
check_overreachAudit a code diff against the originating natural-language prompt. Flags every out-of-scope change the AI agent made — unauthorized deps, env vars, endpoints, cron jobs, files, and features. Returns structured findings + a scope_creep_score.
healthHealth check for the Overreach MCP server.
BEFORE YOU INSTALL
2 tools when we connected. The MCP tool is best-effort self-audit: an agent can skip check_overreach or ignore its findings. Enforcement comes from the CLI and the GitHub Action.
npx on PATH
All 3 documented clients run the server on your machine.
CONTEXT COST
Rated by the size of the 2 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
npx overreach --serveclaude mcp add overreach -- npx -y overreach --serve{
"mcpServers": {
"overreach": {
"command": "npx",
"args": [
"-y",
"overreach",
"--serve"
]
}
}
}MCP servers
jurek-f
johnnywuj81
Log-LogN
gargoyle92