v2.3.0 · MIT · local-first · no AI calls

Is your repo ready for AI coding agents?

agentlint scores any codebase 0–100 on agent-readiness for Claude Code, Cursor, Codex, Copilot, and Gemini CLI — then hands you the fix prompts to close the gap. Local. Free. 30 seconds.

Read the docs →How it compares →Leaderboard →

This is a real report

Generated by running the actual CLI against a typical Express/TypeScript service. Same components, same data shape your dashboard renders after every push.

acme/checkout-apiagentlint v2.3.0

13 pass · 9 fail

discoverability

19/25

buildability

24/25

conventions

3/20

documentation

n/a

safety

8/15

Top findings

fail
tool-shims-present
No tool-specific shims found. Claude Code, Cursor, and Copilot won't see your AGENTS.md content.
fail
linter-config
No linter configuration detected.
fail
formatter-config
No formatter configuration detected.
fail
editorconfig
.editorconfig not found.

fix prompt — paste into any coding agent

Add per-tool shim files so every agent finds the AGENTS.md content. The finding says: "No tool-specific shims found. Claude Code, Cursor, and Copilot won't see your AGENTS.md content."
1. CLAUDE.md at the repo root containing: a one-line note plus `@AGENTS.md` (Claude Code import syntax) or a brief pointer to AGENTS.md.
2. .github/copilot-instructions.md — a pointer to AGENTS.md plus any Copilot-specific notes.
3. .cursor/rules/main.md (or legacy .cursorrules) — same content pointer for Cursor.
Each shim should reference AGENTS.md as the single source of truth, not duplicate it.

How it works

Scan

Run one command in any repo. 30 checks across 5 categories, a 0–100 score, and a fix list — in about 30 seconds, fully local.

npx @agentlinthq/cli@latest .

Fix — with prompts, not homework

Every finding ships a predefined fix prompt. `agentlint prompt` prints one consolidated prompt you paste into Claude Code, Cursor, Copilot, or Codex. The agent does the work; agentlint verifies it.

agentlint prompt | pbcopy

Track & enforce

Push reports from CI to a hosted dashboard: score trends, category charts, public badge, PR comments, and a minimum-score policy that gates merges.

agentlint init && agentlint . --push

How it scores

Five weighted categories, renormalized to 0–100 against what applies to your repo. The full methodology is documented; the rules are open source.

Category	Checks	Weight
Discoverability	README, AGENTS.md, repo metadata	25
Buildability	Lockfile, scripts, env template	25
Convention clarity	Commit format, branch hygiene	20
Documentation surface	License, contributing, code of conduct	15
Safety & guardrails	Secrets, SECURITY.md, dependency hygiene	15

The dashboard: every run, remembered

The CLI stays free and local. Push reports from CI and the hosted dashboard turns them into trends, charts, badges, and merge gates — for $5/month per org, flat.

Score trends

Every push lands as a run. Watch the score move per branch, per project, over time.

Category breakdowns

See exactly which of the five categories is dragging the score, on every run.

Fix prompts in the browser

Each run keeps its prompts. Copy the consolidated fix prompt straight from the run page.

PR comments

The GitHub App posts the score diff on every pull request. No config beyond install.

Public badge

A shields-style SVG badge for your README, score-color-coded, cached at the edge.

Policy gates

Set a minimum score; the CLI exits non-zero in CI when a run falls below it.

See pricing — $5/mo Sign in with GitHub