Built for agents first.
Some employers will never open this site — their agent will. Some builders will have an agent apply for them. That's fine with us: every core action here — search talent, request a shortlist, submit proof, check status — is available over MCP and plain JSON, no auth, no scraping required. Humans review every submission either way.
MCP server
One endpoint, seven tools, the whole hiring loop. Stateless streamable HTTP — point any MCP client at it.
- list_roles — the fractional role spectrum with engagement shapes
- list_engagement_types — the shapes a hiring need can take: build, workflow install, process change, capability mapping, embedded operator
- search_talent / get_talent_profile — vetted builders with public proof
- search_showcases — shipped work, for calibrating a brief or vetting a builder
- submit_talent_brief — request a human-reviewed shortlist (employer side)
- submit_talent_proof — apply to the network (talent side)
- check_submission_status — poll review state by submission id
"mcpServers": { "ai-atlas": { "url": "https://hireainative.com/api/mcp"
} } }
REST API
Everything the MCP server exposes is also plain JSON over GET/POST. No keys, no rate-limit dance for read endpoints.
- GET
/api/roles— role categories, engagement shapes, talent counts- GET
/api/talent?role=ai-engineer&q=evals— search the network;?handle=for one profile with showcases- GET
/api/showcases?q=agents— public proof of shipped AI work- POST
/api/briefs— employer brief:outcome_goal, company_context, current_workflow, urgency, contact_name, contact_email- POST
/api/talent-submissions— talent application:display_name, contact_email, target_outcomes, proof_summary, proof_links, privacy_notes- GET
/api/submissions/:id— poll status of aBRF-…orTLT-…submission
Rules of engagement
Act with a mandate
Only submit briefs or applications with the explicit consent of the person whose contact details you provide. Replies go to that email — make sure they expect it.
Agents submit, humans decide
Every brief and application is reviewed by a person. Volume doesn't help; a precise outcome statement and real proof links do.
Use the API, skip the scrape
All public data is served as JSON with stable shapes. Talent contact details are never exposed — introductions run through briefs by design.
Point your agent at the network.
Start with llms.txt, connect the MCP server, or just curl the API. The same matching process runs no matter who — or what — fills in the form.