{"data":{"slug":"sebuzdugan-frai","name":"frai","tagline":"A toolkit for responsible AI development that generates model cards, risk assessments, and evals via CLI and SDK.","github_url":"https://github.com/sebuzdugan/frai","owner":"sebuzdugan","repo":"frai","owner_avatar_url":"https://avatars.githubusercontent.com/u/81054605?v=4","primary_language":"TypeScript","stars":54,"forks":4,"topics":["ai","ai-compliance","ai-governance","artificial-intelligence","audit","bias","evaluation","frai","guardrails","machine-learning","ml","nlp","open-source","responsible-ai","safety","transparency"],"archived":false,"github_pushed_at":"2026-10-14T16:08:34+00:00","maintenance_label":"Active","stars_delta_30d":1,"url":"https://www.graphcanon.com/tools/sebuzdugan-frai","markdown_url":"https://www.graphcanon.com/tools/sebuzdugan-frai.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/sebuzdugan-frai","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=sebuzdugan-frai","description":"Does your product tell people it uses AI? One command checks your code, live site and spec, and fails the PR until it does. EU AI Act Art. 50.","homepage_url":"https://frai.cc","license":"MIT","open_issues":1,"watchers":0,"ai_summary":"Provides an open-source framework for auditing AI projects to ensure compliance with safety standards by scanning code, generating documentation on model performance risks, and more.","readme_excerpt":"<div align=\"center\">\n\n# FRAI\n\n**Does your product tell people it uses AI?**<br>\nOne command checks your code, your live site and your spec, and fails the pull request until the answer is yes.\n\n\n\n\n\n\n<img src=\"assets/demo.gif\" width=\"820\" alt=\"npx frai on frai.cc's own code: finds OpenAI in 15 files and the notice in chat-client.tsx, confirms the live site, adds the gate, drafts the spec, blocks on 15 questions only a person can answer, then passes CI\">\n\n<sub>A real run on frai.cc's own code. Only step 4 is scripted: a stand-in answers the 15 open questions.</sub>\n\n</div>\n\n## Try it\n\n```bash\nnpx frai                      # review this repo: code, live site, spec\nnpx frai check yoursite.com   # check any site, no repo needed\nnpx frai init                 # add the gate to every pull request\n```\n\nNo key, no signup. Node 18 or newer.\n\n## How it works\n\n\n\n- **Code.** Finds AI SDKs and model calls (OpenAI, Anthropic, Gemini, LangChain and 8 more), then the sentence in your UI that tells people, down to the file and line.\n- **Site.** Finds your URL in `package.json`, `.env` or the README, reads the homepage and policy pages like a visitor, and opens your chat widget to read what it says first.\n- **Spec.** `FRAI-SPEC.md` asks seven questions: risk tier, data, oversight, evaluation, bias, monitoring, transparency. `frai draft` fills them from your code and writes `NEEDS HUMAN INPUT: <question>` wherever the code can't know. The gate blocks until a person answers.\n\n## What it finds in the wild\n\n\n\nMost AI notices live inside the chat, not on the page, so a page scan misses them. FRAI declines cookies, opens the chat through the vendor's own API, reads the first screen and keeps a screenshot. Where it can't read the chat, it says so instead of guessing.\n\n## The report\n\n<img src=\"assets/report.png\" width=\"720\" alt=\"FRAI full report for intercom.com: notice found, with the chat's greeting quoted and a screenshot of the open chat\">\n\n`npx frai check yoursite.com --email you@company.com` (or [frai.cc](https://frai.cc)) sends the full report: up to 10 pages, the chat opened and quoted, a screenshot, the wording to add, and a PDF. Add `--watch` and it re-checks every month. All free.\n\n## Why now\n\nSince 2 August 2026, EU AI Act Article 50 says people must be told when they're talking to an AI or seeing AI-generated content. Fines reach €15M or 3% of global turnover. Model makers watermark their output for machines. Telling the person is your job, and it is the half that breaks quietly when someone swaps a chat widget or adds a second model.\n\n## In CI\n\n```yaml\n- run: npx frai@latest --ci --offline\n```\n\n`frai init` writes this workflow for you. It fails when:\n\n- the spec still has open questions,\n- AI code lands with no spec,\n- the code starts calling an AI provider the spec never names.\n\nDrop `--offline` to also fail when the live site has no notice. The gate is deterministic: no model, no key. Add `--smart` for an AI review of vague answers.\n\n## What leaves your machine\n\n| | Runs where |\n|---|---|\n| Code scan, notice-in-code, gate | On your machine. Nothing is sent. |\n| Site check | Through frai.cc, which visits your public site as [FRAIBot](https://frai.cc/bot) and respects robots.txt. Skip it with `--offline`. |\n| `frai draft` | Your own Claude (Claude Code login or `ANTHROPIC_API_KEY`). Without one, frai.cc drafts it after listing the files it would send and asking you first. |\n\n## Packages\n\n| Package | |\n|---|---|\n| [`frai`](https://www.npmjs.com/package/frai) | The CLI. |\n| [`frai-gate`](https://www.npmjs.com/package/frai-gate) | The gate alone: `npx frai-gate check FRAI-SPEC.md`, or `import { validateSpec } from 'frai-gate'`. |\n| [`frai-core`](https://www.npmjs.com/package/frai-core) | Scanners and document helpers under the CLI. |\n\nBuilding with coding agents? `npx skills add sebuzdugan/frai-skills` teaches them to write specs that include the gate.\n\n## Contribute\n\n```bash\npnpm install\npnpm --filter frai-gate build && pnpm --filter frai buil","github_created_at":"2025-07-13T11:11:18+00:00","created_at":"2026-07-15T10:45:01.998048+00:00","updated_at":"2026-10-23T06:00:15.741576+00:00","categories":[{"slug":"developer-tools","name":"Developer Tools","url":"https://www.graphcanon.com/categories/developer-tools","markdown_url":"https://www.graphcanon.com/categories/developer-tools.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/developer-tools"},{"slug":"evaluation-observability","name":"Evaluation & Observability","url":"https://www.graphcanon.com/categories/evaluation-observability","markdown_url":"https://www.graphcanon.com/categories/evaluation-observability.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/evaluation-observability"}],"tags":[{"slug":"ai-compliance","name":"ai-compliance"},{"slug":"audit","name":"audit"},{"slug":"bias","name":"bias"},{"slug":"evaluation","name":"evaluation"},{"slug":"frai","name":"frai"},{"slug":"machine-learning","name":"machine-learning"},{"slug":"model-card","name":"model card"},{"slug":"risk-assessment","name":"risk-assessment"}],"trust":{"provenance":{"is_fork":false,"github_id":1018927818,"owner_type":"User","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-10-23T06:00:13.566Z","maintenance":{"label":"Active","score":82,"methodology":"github_public_v1","releases_90d":0,"days_since_push":8,"last_release_at":null,"stars_delta_30d":1,"open_issues_delta_30d":1},"security_summary":{"last_scan_at":"2026-07-15T10:45:04.903289+00:00","scan_profile":"deps","status":"ok","critical_count":0,"high_count":0,"medium_count":0,"low_count":0,"scanner":"osv@v1","scan_id":"c371f4df-2637-404b-82af-fd47369ab5d7","pending_review_count":0}},"capability_facts":{"mcp":{"method":"github-tree-and-contents@2","reason":"No explicit MCP server declaration was found in the inspected manifests. Other layouts and implementations were not tested.","source":"repo_scan","status":"unknown","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.122Z","server_manifest":null},"scan":{"method":"github-tree-and-contents@2","source":"repo_scan","status":"complete","version":2,"coverage":{"scope":"bounded-repository-manifests","tree_complete":true,"selected_files":5,"candidate_files":5},"revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.122Z","observations":[{"path":"@revision","bytes":3439,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/commit/f563075b3d9faa93fa6d53382b7501fe199d9bca","http_status":200,"observed_at":"2026-10-23T06:00:14.471Z"},{"path":"@tree","bytes":45702,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","http_status":200,"observed_at":"2026-10-23T06:00:14.611Z"},{"path":"docker-compose.yml","reason":"not-listed-in-complete-tree","status":"not-found","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.611Z"},{"path":"Dockerfile","reason":"not-listed-in-complete-tree","status":"not-found","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.611Z"},{"path":"mcp.json","reason":"not-listed-in-complete-tree","status":"not-found","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.611Z"},{"path":"package.json","bytes":731,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/blob/f563075b3d9faa93fa6d53382b7501fe199d9bca/package.json","http_status":200,"observed_at":"2026-10-23T06:00:14.734Z"},{"path":"packages/frai-agent/package.json","bytes":1004,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/blob/f563075b3d9faa93fa6d53382b7501fe199d9bca/packages/frai-agent/package.json","http_status":200,"observed_at":"2026-10-23T06:00:14.736Z"},{"path":"packages/frai-cli/package.json","bytes":1133,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/blob/f563075b3d9faa93fa6d53382b7501fe199d9bca/packages/frai-cli/package.json","http_status":200,"observed_at":"2026-10-23T06:00:14.763Z"},{"path":"packages/frai-core/package.json","bytes":784,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/blob/f563075b3d9faa93fa6d53382b7501fe199d9bca/packages/frai-core/package.json","http_status":200,"observed_at":"2026-10-23T06:00:14.777Z"},{"path":"packages/frai-gate/package.json","bytes":1311,"status":"present","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/blob/f563075b3d9faa93fa6d53382b7501fe199d9bca/packages/frai-gate/package.json","http_status":200,"observed_at":"2026-10-23T06:00:14.911Z"},{"path":"pyproject.toml","reason":"not-listed-in-complete-tree","status":"not-found","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.611Z"},{"path":"server.json","reason":"not-listed-in-complete-tree","status":"not-found","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.611Z"}]},"has_cli":{"value":true,"method":"repository-declaration@2","source":"packages/frai-cli/package.json:bin","status":"declared","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/blob/f563075b3d9faa93fa6d53382b7501fe199d9bca/packages/frai-cli/package.json","observed_at":"2026-10-23T06:00:14.122Z"},"languages":{"value":["typescript","javascript"],"method":"github-tree-and-contents@2","source":"github.language+package.json+packages/frai-agent/package.json+packages/frai-cli/package.json+packages/frai-core/package.json+packages/frai-gate/package.json","revision":"f563075b3d9faa93fa6d53382b7501fe199d9bca","source_url":"https://github.com/sebuzdugan/frai/tree/f563075b3d9faa93fa6d53382b7501fe199d9bca","observed_at":"2026-10-23T06:00:14.122Z"},"license_spdx":{"value":"MIT","source":"github.license","observed_at":"2026-10-23T06:00:14.122Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":null,"constraints":null,"when_to_use":["Use when you need to comply with specific safety standards in AI projects using JavaScript framework.","Preferred when integrating automated evaluation processes into your workflow via CLI commands."],"when_not_to_use":["Avoid if your development environment strictly requires non-JavaScript languages for tool integration.","Not advisable if automated governance and scanning tools are not aligned with organizational compliance policies."],"source":"enrich:decision_facts","observed_at":"2026-07-16T19:34:13.458Z"},"constraint_facets":null,"decision_summary":[{"label":"Adopt for","value":"Unverified catalog guidance: frai is an open-source toolkit that uses CLI and SDK to scan code for responsible AI development purposes, generating model cards, risk assessments, and evaluations."}]}}