
Security News
Lovable’s OJ Rewrites Vite’s Dev Server in Rust as AI Lowers the Cost of Forking Open Source
Lovable’s OJ rewrites Vite’s dev server in Rust, reducing memory use and preview times as AI lowers the cost of open source reimplementation.
verdict-qa-mcp
Advanced tools
Read-only MCP server and release gate over Verdict QA state: verdicts, findings, flaky quarantine, run history.
A read-only MCP server and release gate over Verdict's QA state — so anything that speaks MCP can consult your QA memory.
Verdict is a Claude Code QA agent that keeps a baseline and reports what broke since
yesterday: findings with stable IDs and ages, every red test classified, flaky tests
quarantined with an expiry, and a verdict of pass | pass with risks | blocked | fail
that names what was not tested.
This distribution is the part of Verdict that other tools talk to. It reads the same state files the agent writes and never writes to them — an orchestrator gating a merge, a Cursor or Codex session, or a CI step commenting on a PR can all ask what the tester last found. The agent itself is a Claude Code plugin with zero dependencies and works without this package.
Note on the name. The distribution is
verdict-qa-mcp; the console script is stillverdict-mcp, and the import package is stillverdict_mcp.verdict-mcpwas taken on PyPI by an unrelated project.
claude mcp add verdict -- uvx --from verdict-qa-mcp verdict-mcp
Or pip install verdict-qa-mcp / uv pip install verdict-qa-mcp.
Every tool carries a read-only annotation, and the server never writes — the tester's memory is public API; the tester's pen is not.
| Tool | Returns |
|---|---|
get_verdict(project) | last verdict, release blockers, report path, not-tested list |
get_findings(project, status) | open (default), all, or NEW / STILL_OPEN / RESOLVED / REGRESSED — REGRESSED ranked first |
get_quarantine(project) | the flaky ledger, each entry with a computed expired flag |
get_history(project) | run-over-run trend parsed from the report index |
get_report(project, report?) | full report content, path-guarded to the QA root — so CI can quote the evidence, not just link it |
get_profile(project) | isolation rules, risk areas, real test commands, and the lessons ledger when one exists |
get_trends(project) | trajectory, current pressure (open by severity, age, quarantine size), and hotspots — where this project's defects actually cluster |
list_projects() / get_state(project) | everything with a baseline / the raw state |
project is a key under the solo root (~/.claude/verdict/, override with VERDICT_HOME)
or a repo path in team mode, which resolves <repo>/.qa/.
| Command | Does |
|---|---|
verdict-mcp | the MCP server above |
verdict-gate | exit-code release gate for CI — keeps "never ran" distinct from "said no" |
verdict-validate | checks a state file is well-formed (--at-rest for a committed one) |
verdict-run | runs a pass through the harness; --skip-unless-drift carries the standing verdict with no model call when nothing a finding cites has moved, and --on-drift local hands a blocked sweep to verdict-local instead of spending a Claude run |
verdict-local | a pass driven by the harness, with a small local model answering one bounded question at a time (an Anthropic-compatible gateway in front of Ollama, for example). --delta carries every prior open finding by measurement, by id, or by re-filing it — never by silence; --range/--base with a throwaway --qa-root judge a branch without touching the project's own state |
verdict-accept | the maintainer's pen: record that a finding's risk is accepted, with a citation and a reason — the one status the tester cannot write |
verdict-answer | the maintainer's other pen: answer or dismiss a question the tester parked |
verdict-issues | open findings as GitHub issues through your own gh login — a dry run unless --create |
verdict-facts / verdict-finalize | measure-then-judge harness halves |
Full documentation, the plugin itself, and the published eval results live on GitHub:
MIT licensed.
FAQs
Read-only MCP server and release gate over Verdict QA state: verdicts, findings, flaky quarantine, run history.
We found that verdict-qa-mcp demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.

Security News
Lovable’s OJ rewrites Vite’s dev server in Rust, reducing memory use and preview times as AI lowers the cost of open source reimplementation.

Security News
It has been one year since Shai-Hulud made its first appearance on npm.

Research
/Security News
Operators behind PolinRider used a compromised GitHub account to plant malware in four development versions of a Packagist package with 700,000+ downloads.