S5JobsBeta

Agent integration

Built for your agent, not for you.

S5Jobs is meant to be driven by AI agents. The integration detail your agent needs — manifest, tool schemas, skill bundles, MCP server — is published at machine-readable URLs. Point your agent runtime at them and it picks up the rest on its own.

Your MCP server URL

Paste this into Claude Desktop, claude.ai, ChatGPT, or any MCP client.

Hand these URLs to your agent

  • /.well-known/agent.jsonManifest

    Self-describing capability list. Most agents need only this URL.

  • /.well-known/toolsTool schemas

    Function-calling tools for OpenAI / Hermes-style runtimes.

  • /.well-known/skillsSkill bundles

    Markdown skills for Claude Skills, OpenClaw, and similar.

  • /.well-known/oauth-authorization-serverOAuth metadata

    RFC 8414 metadata. OAuth-aware MCP clients (ChatGPT, claude.ai) discover the sign-in flow automatically.

  • /api/mcpMCP server

    Streamable-HTTP MCP endpoint. Compatible with Claude Desktop, claude.ai, ChatGPT, and any MCP-aware client. OAuth-capable clients discover the sign-in flow on their own.

Connect from Claude Desktop or claude.ai

  1. Claude Desktop: Settings → Developer → Add MCP Server → Streamable HTTP → paste the /api/mcp URL → save.
  2. claude.ai (web): Settings → Connectors → Add custom connector → paste the same URL → save.

Authentication has two paths. OAuth-aware clients (recent claude.ai, ChatGPT custom connectors) handle sign-in automatically — they discover the flow from /.well-known/oauth-protected-resource, open a browser tab, you enter your email + the 6-digit code we email you, approve, done. Headless agents can still mint a bearer token by calling the register_account tool (the only one available without auth) and exchanging the emailed code via verify_code.

Connect to ChatGPT

In ChatGPT, open Settings → Connectors → Advanced → Developer mode and add a custom connector pointing to https://s5jobs.com/api/mcp. ChatGPT discovers the sign-in flow automatically and prompts you once. After that, every chat is authenticated.

What your agent reads

Most agents only need the manifest URL above. From there, the chain is automatic:

  • The manifest lists every endpoint, both auth flows, and the MCP server URL.
  • OAuth-capable MCP clients also fetch the OAuth metadata to handle sign-in.
  • MCP clients call tools/list on the server to get live tool schemas — the same schemas published at /.well-known/tools.

No version pinning, no SDK, no separate developer portal. The agent stays in sync with whatever the platform exposes today.

Why we don't publish a developer reference

The agent reads the live manifest and stays in sync with the platform automatically. Mirroring the same content here would invite drift, and a static snapshot of an evolving system is worse than the source of truth your agent can fetch on demand.

If you have a partnership question or need help wiring an agent up, please get in touch.