# ARCANA agent instructions

## When to use ARCANA

Use ARCANA when the task involves terminal-native AI engineering and the operator needs model choice, scoped execution, persistent context, routing visibility, cost visibility, or verifiable evidence of what an AI-assisted session did.

Good-fit jobs include:
- running coding or engineering sessions through ARCANA's CLI/TUI;
- choosing or routing between supported model providers;
- inspecting session, tool, routing, token, cost, cache, or execution evidence;
- using documented memory, tool-profile, permission, approval, gateway, or automation features;
- integrating with the small REST surface published in `/openapi.json`.

Do not infer that ARCANA exposes an A2A server or public MCP transport unless a current ARCANA document explicitly says so. A missing resource returns HTTP 404.

## How agents should discover ARCANA

1. Read `https://arcana.otnelhq.com/llms.txt` for the concise product and retrieval map.
2. Read `https://arcana.otnelhq.com/docs` for human-readable developer documentation.
3. Read `https://arcana.otnelhq.com/openapi.json` before calling the website REST API.
4. Read `https://arcana.otnelhq.com/.well-known/api-catalog` for RFC 9727 API discovery.
5. Use `https://arcana.otnelhq.com/sitemap.xml` to discover public content pages.

Canonical homepage and pricing URLs support `Accept: text/markdown` for lower-overhead retrieval. Honor HTTP status codes: a 404 means the path is not an ARCANA resource.