How to use FrootAI with Cursor, Windsurf, and Claude Code
Your AI coding assistant already knows how to write code. Now teach it how to build AI agents.
Install one rule pack and your assistant becomes a FrootAI expert — it knows the FAI Protocol, suggests correct manifest structures, generates eval datasets, and links you to Studio for visual editing.
Cursor
Add the rule file to your project:
``bash
`Copy the .mdc file to your .cursor/rules/ directory
curl -o .cursor/rules/frootai.mdc \
https://raw.githubusercontent.com/frootai/frootai/main/community-plugins/cursor-rules/frootai.mdc
That's it. Now when you work with .fai-manifest.json files, Cursor knows:
- All 9 primitive types and 4 required fields
- How to structure eval datasets (JSONL, ≥ 20 cases)
- Studio canvas commands and GitHub sync
- Debugging tips (groundedness < 0.8 = hallucinating)
Windsurf
`bash
`Copy to your .windsurf/rules/ directory
curl -o .windsurf/rules/frootai.yml \
https://raw.githubusercontent.com/frootai/frootai/main/community-plugins/windsurf-rules/frootai.yml
Claude Code
`bash
`Copy to your project root or .claude/ directory
curl -o CLAUDE.md \
https://raw.githubusercontent.com/frootai/frootai/main/community-plugins/claude-code-rules/CLAUDE.md
Or add it to your existing CLAUDE.md by appending the content.
What the rule pack teaches your assistant
| Capability | What it does |
|---|---|
| Manifest structure | Knows all 4 required fields, 9 primitive types, semver versioning |
| Eval datasets | Generates JSONL with edge cases (empty, long, adversarial, multilingual) |
| Studio integration | Adds x_studio.canvas.nodes[].pos for visual layout |
| GitHub sync | Suggests .frootai/eval-config.json and CI eval action |
| Debugging | Interprets eval scores (groundedness, helpfulness, safety) |
| WAF pillars | Aligns plays with Well-Architected Framework |
Bonus: MCP server
For even deeper integration, install the FrootAI MCP server:
`bash
npx frootai-mcp
`
This gives your AI assistant access to FrootAI tools:
- fai.validate — validate a manifest
- fai.search — search plays and primitives
- fai.eval — run eval on a manifest
Try it now
1. Install the rule pack for your tool (above) 2. Open a .fai-manifest.json` file 3. Ask your assistant: "Create an Enterprise RAG play with eval hooks" 4. Watch it generate a valid manifest with all the right structure
The AI coding assistant becomes your fastest path to building AI agents.
— Pavleen