Developer Support

Supercharge your AI in the real world.

Find answers, troubleshoot issues, and get in touch with the RakuAI team. Six-tool MCP surface, 9,500+ REST endpoints, multi-vendor by design — Claude, ChatGPT, Gemini, Copilot.

Frequently Asked Questions

How do I get an API key? ▼

Accept the developer NDA, then go to the dashboard to create API keys. You can create up to 5 active keys with different scopes (read, generate, full).

Which AI clients can drive the runtime? ▼

Any MCP-capable client. Today: Claude Desktop (full wiring guide and PowerShell auto-config in /developers/claude-desktop.html). In flight: Cursor, GitHub Copilot in agent mode, custom GPTs over the MCP HTTP transport. The six tools (load_world_model, ingest_frame, get_scene_state, set_render_target, start_simulation, get_metrics) are the same contract for every client.

What is the MCP surface? ▼

Six tools exposed over Model Context Protocol: load_world_model, ingest_frame, get_scene_state, set_render_target, start_simulation, get_metrics. Read-only tools work in every environment; mutating tools are sandbox/dev only, deny-by-default. The server enforces typed schemas and writes a full audit log. See the build log for the full reasoning.

What are the rate limits? ▼

Free tier: 10 generations/day, 30 iterations/day. Pro tier: 100 generations/day, 500 iterations/day. Enterprise: unlimited. Exceeded limits return HTTP 429 with a Retry-After header. See rate limits documentation.

Can I use RakuAI in a commercial product? ▼

Yes. Integrations built on the MCP surface can ship commercially. The SDK and runtime are covered by the developer NDA, and the Raku runtime engine requires a valid license for distribution. Contact us for licensing details.

How is AI involvement disclosed? ▼

The runtime records every MCP tool call in an audit log: caller, tool, arguments, timestamp, outcome. Adapter manifests declare which world-model backend (video predictor, neural radiance field, physics foundation model, etc.) is in use. The engine retains deterministic authority on physics, collision, and scoring regardless of which adapter is contributing frames.

Troubleshooting

401 Unauthorized

Your API key or JWT token is missing, expired, or invalid.

Check your Authorization: Bearer <token> header. Regenerate a key from the dashboard if needed.

403 Developer Access Required

You're trying to access developer endpoints without NDA acceptance.

Visit the registration page to accept the developer NDA first.

422 Validation Error

The request body is missing required fields or has invalid values, or an MCP tool call failed schema validation at the server.

Check the response detail field for specifics. For MCP tool calls, confirm argument types match the tool schema in the Claude Desktop guide.

429 Too Many Requests

You've exceeded your tier's rate limit.

Check the Retry-After header for when to retry. Upgrade your tier for higher limits.

503 Service Unavailable

The runtime or an upstream world-model adapter is temporarily unavailable.

Retry after a few seconds. Read-only tools (get_metrics, get_scene_state) recover first; check those before mutating calls.

Community

💻

GitHub

Source code, issues, discussions

💬

Discord

Real-time chat with developers

📢

Updates and announcements

Contact Us

Name