Agent resources
Give an agent one setup file, one documentation page, or the whole corpus. Public, no key, clean Markdown.
Every endpoint on this page is public, requires no account, and returns Markdown rather than an HTML page an agent has to strip.
Agent onboarding#
/agents.md tells an agent how to find a key, configure the base URL, send a real request, and recover from the errors it is most likely to hit. It is the one file to hand an agent that has never seen OneRouter.
One page at a time#
Append .md to any docs URL:
curl https://onerouter.dev/docs/models.mdOr ask for it by content type, using the normal URL:
curl -H "Accept: text/markdown" https://onerouter.dev/docs/modelsIndex, or the whole corpus#
# Short index: fetch only the pages you need
curl https://onerouter.dev/llms.txt
# Every documentation page in one response
curl https://onerouter.dev/llms-full.txtUse the index when the agent can fetch individual pages on demand. Use the full file when a tool needs the entire corpus in a single request and you would rather spend context than round trips.
Documentation MCP server#
Point a Streamable HTTP MCP client at the public endpoint:
https://onerouter.dev/mcp
Tools: search_docs, get_doc
Resources: every documentation pageAll operations are read-only. Server metadata is at /.well-known/mcp/server-card.json.
Agent skill#
The reusable skill lives at /skills/onerouter/SKILL.md, discoverable from /.well-known/agent-skills/index.json.
Machine-readable catalogs#
| Endpoint | Contents |
|---|---|
https://onerouter.network/v1/catalog | Every model with context window, modalities and current price |
/api/v1/auto | The live onerouter/auto resolution table, per intent |
/api/v1/status | Host health, per model family |
They change as capacity does. An agent that reads /api/v1/auto before a run gets the picks that are live now; one that hardcoded last month's picks gets auto_unresolvable and deserves to.