Stops your AI coder from silently re-introducing bugs, leaking provider keys, or weakening tests. Pinned turns each bug…
Stops your AI coder from silently re-introducing bugs, leaking provider keys, or weakening tests. Pinned turns each bug fix into a permanent regression test, and blocks AI commits that try to skip / weaken / delete tests in `tests/pinned/`. **Six tools your AI agent can call:** - `pinned_before_code_change` — surfaces active guards + AI lessons in the repo BEFORE the agent edits anything; recommends `npx pinnedai init` if the repo is unprotected - `pinned_before_done_check` — full pre-completion guard run; returns `status = pass | review | block`. Agents must report the result to the user before saying work is done - `pinned_scan_diff` — finds risky surfaces in the current diff (new routes, webhooks, env edits, middleware) that don't yet have a guard - `pinned_list_guards` — lists active Pinned guards so the agent knows what behavior is protected - `pinned_check_pr_description` — parses PR bodies / commit messages for behavioral claims (auth, rate-limits, idempotency, permissions) and structures them - `pinned_suggest_init` — detects whether the repo has Pinned set up and recommends `npx pinnedai init` if not Every response includes a `human_summary` and `agent_instruction` field so the agent reports guard status in its final answer rather than using Pinned silently. Local-first. No telemetry. Pairs with the `pinnedai` CLI (`npm install -g pinnedai`). Apache-2.0.
michaelzon7
mcp
free
Others in the same category, ranked by how often they are opened.
