agentsky.dev publishes plain-text files written for AI agents, so an agent can learn AgentSky and get set up without you explaining it.
Give your agent one of these addresses and let it read from there:
https://agentsky.dev/start.md— agent self-onboarding. It explains what AgentSky is, how to get access without you typing secrets into the chat (sky auth login --no-browserprints a link for you to approve, or you can provide an API key), the two MCP servers, how to make a first API call, and the help search API. It also says where/setup.mdand/onboard.mdfit.https://agentsky.dev/llms.txt— a short, curated index of the site. It opens with a "For AI agents: start here" section linking to/start.md,/setup.mdand/onboard.md, then lists the product, agents, developer docs and Help Center.https://agentsky.dev/llms-full.txt— the full listing of model, tool, agent and connector pages, for agents that need everything./llms.txtlinks to it.
Machine-readable discovery files:
https://agentsky.dev/.well-known/mcp.jsonlists both MCP servers: the developer serveragentskyathttps://api.agentsky.dev/v1/mcp(needs an API key as the bearer token) and the keyless Help Center serveragentsky_helpathttps://api.agentsky.dev/v1/help/mcpwith the toolshelp_searchandhelp_get.https://agentsky.dev/openapi.jsonandhttps://agentsky.dev/.well-known/openapi.jsonredirect to the API spec athttps://api.agentsky.dev/v1/openapi.json.https://agentsky.dev/.well-known/security.txtgives the contact for reporting security issues.
Where: agentsky.dev — any agent that can fetch a web page
Good to know: You don't need an account to read these files. Related: Use AgentSky as an MCP server, Use the AgentSky API, Create and manage API keys, Use the sky CLI.
