AgentNavaKit Docs
Ship an AI agent without writing the runtime.
AgentNavaKit is a managed agent runtime. Describe an agent declaratively with defineAgent; the runtime executes it and emits a typed event stream any UI can render. Bring your own UI — React, Vue, Svelte, vanilla, a CLI, an embed widget.
Vibe-coding? Skip the docs.
Paste this prompt into Claude Code, Codex, or Cursor. It installs AgentNavaKit, scaffolds a working agent, and runs it locally — without you reading a line.
Set up AgentNavaKit (`@agentnava/kit`) in this project end-to-end.
1. Detect package manager from the lockfile (bun.lockb → bun, pnpm-lock.yaml → pnpm, yarn.lock → yarn, else npm). Use it consistently.
2. Install `@agentnava/kit` and `valibot`.
3. Create `agents/hello.ts` with a working `defineAgent({...})` spec — modelClass: 'standard', chat trigger, helpful concise systemPrompt.
4. Add scripts: `"agent:dev": "agentnava-kit dev agents/hello.ts"` and `"agent:deploy": "agentnava-kit deploy agents/hello.ts"`.
5. Run `agent:dev` and confirm it boots on http://localhost:8787. If port is occupied, ask before changing.
6. Curl-test the chat endpoint and show me the streaming SSE response.
7. Print a 1-paragraph summary of what was installed and what to try next.
Reference: https://docs.agentnava.com/docs/getting-started
Don't modify unrelated files. Ask before doing anything ambiguous.