nspection
Inspection, inside out

Get a trust grade for your
AI agent & code

nspection inspects any GitHub repo across 6 security dimensions and hands you a shareable, deterministic report card. No signup. No black boxes.

Inspect a repo β†’
$npx nspection .

What we grade

Six dimensions covering the full attack surface of AI-era codebases. Same repo always gets the same grade.

πŸ”“

Secret Hygiene

Leaked keys, tokens, API secrets in source, configs, .env files, and git history.

deterministic
🧠

Prompt-Injection Defense

Plaintext system prompts, eval(), remote instruction fetches, tool-call injection surfaces.

AI-specific
πŸ€–

Agent Autonomy Safety

Agent frameworks present + dangerous tool surfaces: exec, shell, SQL, file write, network.

AI-specific
πŸ“¦

Supply-Chain Risk

Dependency CVEs, abandoned or typosquat packages, lockfile integrity, pinning hygiene.

deterministic
βš–οΈ

License Clarity

Copyleft contamination risk, permissive vs restrictive licensing, unknown/missing license flags.

deterministic
πŸ§ͺ

Guardrail & Eval Coverage

Test suites, CI pipelines, moderation/safety code, eval harnesses, and observability mechanisms.

deterministic

Platform services

Everything you need to inspect, grade, and share trust reports β€” all from one platform.

πŸ›°οΈ

Repo Inspector

Point it at any public GitHub repo. We clone, scan, and grade it across 6 dimensions. Deterministic β€” same code always gets the same grade. No LLM required.

Try it now β†’
🎴

Shareable Report Cards

Every inspection generates a 1200Γ—630 PNG card β€” optimized for OG/Twitter embeds. Post your grade anywhere; the card tells the whole story at a glance.

See example card β†’
πŸ†

Public Trust Leaderboard

Every inspected repo is ranked by overall trust score. The leaderboard updates live β€” compete for the top spot or find the safest repos to depend on.

View leaderboard β†’
πŸ›‘οΈ

Built-in Guardrails

The platform eats its own dogfood: input validation, target allowlists, malicious-pattern denylists, and per-IP rate limiting protect the public API.

Active on every request
⚑

REST API

Full programmatic access β€” inspect repos, fetch reports, pull leaderboard data, and serve OG cards. Build bots, CI gates, or dashboards on top.

View endpoints β†’
πŸ’»

CLI Tool

Inspect any local or GitHub repo from your terminal. Coming soon: npx nspection . β€” one command, one grade, zero friction.

Coming soon

How it works

Three steps from zero to a trust grade.

Point at a repo

Paste a GitHub URL or owner/repo. No account needed.

We scan locally

We clone the repo and run deterministic analyzers across all 6 dimensions. No code leaves the platform.

Get your grade

Letter grade, numeric score, dimension breakdown, and a shareable PNG card β€” instantly.

πŸ›°οΈ Run an inspection

Try langchain-ai/langchain, microsoft/autogen, or openai/openai-python. No signup, no upload.

API

Full programmatic access. Inspect, fetch, and share β€” all via HTTP.

POST/api/inspect

Inspect a GitHub repo or local path. Body: {"repo":"owner/name"}. Returns full report with grade, score, and dimensions.

GET/api/leaderboard

Top inspected repos ranked by trust score. Query: ?limit=20. Returns id, repo, grade, score, color.

GET/card/:id.png

Shareable 1200Γ—630 PNG report card for any inspection. Perfect for OG/Twitter embeds.

GET/api/report/:id

Full JSON report for a specific inspection by ID. All 6 dimension scores, metadata, and letter grade.

$curl -s -X POST https://nspection.com/api/inspect -H 'Content-Type: application/json' -d '{"repo":"owner/name"}'

πŸ† Trust leaderboard

View full board β†’
#RepositoryGradeScore
loading…