repograde-mcp
Grades public GitHub repositories (owner/repo) across 33 checks in 6 weighted categories — documentation, community, engineering, discoverability, release hygiene, and agent-readiness. Provides tools to run a full scan and return the overall score, letter grade, per-category scores, pass/warn/fail results, and prioritized fixes, plus a listing of all available checks. Uses the GitHub API to read repo metadata and files, with optional token authentication for higher rate limits.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@repograde-mcpgrade facebook/react and show me the top fixes"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
▲ RepoGrade
Grade your repo before you announce it. 33 checks across 6 categories on any public GitHub repository — in one command, in ~15 seconds, with zero dependencies.
Part of FastPress Labs — small, free, open source tools for people who ship. Sister project: LaunchGrade audits the landing page; RepoGrade audits everything behind the link.
Install
npx -p github:fastpresspages/repograde repograde owner/repoAn npm publish is planned once publishing credentials exist; the GitHub install
path above works today with zero install footprint.
repograde owner/repo # grade any public repo
repograde owner/repo -o report # + report.html/json/md, badge.svg, fix-plan.md
repograde --list # all 33 checks, grouped
repograde --fail-under 80 owner/repo # CI gate (exit 2 when under)Related MCP server: mcp-github-tools
What it grades
Category | Weight | Highlights |
Documentation | 20 | README that explains the repo in 15s, install command, usage example, real changelog |
Community | 15 | LICENSE (permissive preferred), CONTRIBUTING, Code of Conduct, SUPPORT, issue/PR templates |
Engineering | 25 | Tests exist and CI runs them, declared toolchain, lockfile, zero-dep gentle flag |
Discoverability | 15 | Description (15–90 chars), ≥3 topics, homepage link, activity within 90 days |
Release hygiene | 10 | Tagged release, readable notes, tag matches package.json |
Agent-readiness | 15 | AGENTS.md, installable skill, MCP/lang manifest, llms.txt |
Every check is a pure function over the shaped repo context (see src/checks.js)
— the whole scan is deterministic and fixable in PR-sized chunks.
Real example (not marketing)
fastpresspages/launchgrade scanned on 24 October 2026 (v0.1.0):
┌──────────────────────────────────────────────┐
│ A+ 96/100 Announce-ready │
└──────────────────────────────────────────────┘
28 pass · 3 warn · 0 fail
Documentation 100 ████████████████████
Community 100 ████████████████████
Engineering 96 ███████████████████░
Discoverability 100 ████████████████████
Release hygiene 85 █████████████████░░░
Agent-readiness 90 ██████████████████░░
Fix these first
1. WARN [Release] Tag matches package.json
package.json 0.3.0 vs release tag v0.1.0
2. WARN [Agent] llms.txt published
No llms.txt at the repo root
3. WARN [Eng] Dependencies pinned
No lockfile ...Full artifacts for that scan and a lower-grade example live in samples/
(launchgrade → A+ 96, fastpress-start → B 79, same day, honest numbers).
Low grades get published too — that is the point.
The RepoGrade site itself is audited by LaunchGrade: A+ 100/100 at launch (24 Oct 2026, samples/repograde-site/, one intentional no-analytics warning).
Reports & artifacts
Artifact | What it is |
terminal output | colored grade card, per-category bars, top fixes |
| standalone dark report (works offline, shareable) |
| drop into a README or PR |
| machine-readable; |
| repograde badge for your README |
| prioritized fixes with evidence, ready to work through |
GitHub Action
name: RepoGrade
on:
schedule: [{ cron: "0 8 * * 1" }]
workflow_dispatch:
jobs:
grade:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v7
- uses: fastpresspages/repograde@main
with:
fail-under: '75'
env:
GITHUB_TOKEN: ${{ github.token }} # 5,000 req/h instead of 60Outputs score and grade; add --fail-under semantics via fail-under.
MCP server (agents can grade repos too)
{ "mcpServers": { "repograde": {
"command": "npx",
"args": ["-p", "github:fastpresspages/repograde", "repograde-mcp"] } } }Tools: grade_repo (owner/repo → grade + category scores + fixes), list_checks.
MCP check-only rule for agents: batch sanely; the unauthenticated GitHub API allows 60 requests/hour and we respect that limit.
Agent skill
npx skills add fastpresspages/repogradeScoring
Each category score = share of its weight earned (pass=100%, warn=50%, fail=0%).
Overall = weighted mean over scored categories; skip/info checks leave the pool.
Letter grades: A+ ≥95 · A ≥90 · A- ≥85 · B+ ≥80 … F <40.
It is an heuristic about readiness to be announced, not a security audit and not a code review (see SECURITY.md and terms).
Free human review
Open a discussion with your repo + what you are stuck on, drop your repo at the site form, or email meetcode99@gmail.com: we rerun the scan, read the repo and reply with the three changes that matter most — free, no drip campaign. Deeper help (fix sprints, launch builds) exists and is transparently priced in the FastPress Labs playbook.
Principles
Zero dependencies — Node stdlib only.
Public data only — never writes to your repo, never uploads anything.
Publish low grades too.
Fix text a tired human can act on.
No fake metrics: downloads, stars, testimonials — all hand-countable or absent.
Contributing / security / license
CONTRIBUTING.md (check contract, PRs)
SECURITY.md (private disclosure)
Apache-2.0 — see LICENSE
Built by FastPress Labs — see our org profile for everything we ship.
This server cannot be deployed
Maintenance
Related MCP Connectors
Screens public GitHub repos and PRs to generate risk maps, findings, and merge-readiness signals.
Evaluates any GitHub repository for license risk, maintenance velocity, stars/forks, and security.
Evaluates any GitHub repository for license risk, maintenance velocity, stars/forks, and supply-c.
Read-only scan of a public GitHub repo for leaked secrets, with a 0-100 grade.
Related MCP Servers
- AlicenseBqualityCmaintenanceEnables AI agents to interact with GitHub (search repos, read files, issues, PRs), analyze code for quality and issues, and manage tasks with priority sorting.7MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to perform GitHub analytics and management tasks, including repository analysis, PR summarization, issue triage, release notes generation, and contributor statistics.15 npmMIT
- AlicenseAqualityAmaintenanceEnables AI assistants to monitor GitHub repository health by calculating health scores, fetching repository metadata, analyzing Dependabot alerts, checking CI/CD status, and retrieving code scanning alerts.8270 npm1MIT
- AlicenseAqualityBmaintenanceEnables AI agents to analyze public GitHub repositories directly through tools for repo health scoring, recent commit summaries, good first issue discovery, and semantic Q&A over docs and issues.423 npm1MIT