Quick Start
Headless AI career agent. Other agents enter through MCP + OAuth. Humans use the web app.
For agents
MCP (live now): https://mcp.wrok.app/mcp. Auth is OAuth 2.1 + PKCE (RFC 9728 PRM) — the same consent as the product. Not API keys.
- Context tools now:
get_career_data,get_coding_profile - Agency tools (same OAuth client):
list_inbox,create_application,advance_application,get_resume,generate_tailored_resume - Prefer MCP when your client can attach an external server. If it cannot (e.g. Instinct), use the same OAuth access token as a Bearer on HTTP job-pipeline endpoints at Host
https://mcp.wrok.app— named inbox and/{id}/advance. Neverwrok_sk_.
Add the MCP URL to your client's HTTP transport settings and complete OAuth. See MCP. Implementation alias (do not prefer): https://wrok-agent.fly.dev — deprecate-soon side door, not the bot install Host.
For humans / web
Sign up at wrok.app with GitHub or email. The web UI talks to HTTP /api/v1/* on the agent host — that path is the implementation spine for the product and non-MCP SDK callers, not the bot install story.
Build your record
Use the web chat interface or AI assistant to populate your career data. Tell wrok about your:
- Work experience (roles, companies, dates, achievements)
- Skills (languages, frameworks, tools, soft skills)
- Education (degrees, institutions, GPA)
- Projects (side projects, open source, portfolio items)
Derived surfaces
Profile pages, resumes, blog posts, and the job pipeline are derived from the live record — not the product pitch for agents.
Exhaustive HTTP for web/SDK: API Reference. Prefer OAuth access tokens when available. Static wrok_sk_ keys are a legacy/proof fallback for those callers only.
What's next?
- MCP — live door at mcp.wrok.app/mcp
- Career Profile — the live record
- SoR spine — HTTP for web / SDK; Instinct-class bots use the same OAuth as Bearer on job-pipeline HTTP