MCP-native · Agent-first · Open data

Data built for agents to actually use.

Three things, all one MCP call away: custom datasets anyone can create and query, 24k+ companies scored for agent readiness, and Vetted AI researchers, engineers, founders with person-AR scores and recruiter-ready tooling.

13+
Public datasets · co + people
24,000+
Companies · AR scored
AI Builders · people
80+
MCP tools
01 · Datasets

Build a dataset in one MCP call. Query it from anywhere.

Upload a CSV, any schema. We auto-embed every row for semantic search, push updates via HMAC-signed webhooks, and expose keyword + per-column + natural-language queries through a single MCP surface your agents already know how to call.

  • create_custom_dataset — any shape, any schema, live in seconds
  • query_custom_dataset — keyword ILIKE, per-column filters, all server-side
  • semantic_search_dataset — pgvector cosine, scoped to your dataset
  • Webhook-subscribe to row_updated / new_claim — push, not poll
See the dataset MCP tools →
MCP call · one-shot
// Create a dataset from any CSV, get a stable id back
Use OnlyData: create_custom_dataset({
  name: "Sturdy top-50 prospects",
  csv:  "name,domain,segment,...\n..."
})
// Query it three ways:
Use OnlyData: query_custom_dataset({
  dataset_id: "...",
  search:  "fintech boise",
  filters: { segment: "financial_services" }
})
Use OnlyData: semantic_search_dataset({
  dataset_id: "...",
  query: "local banks with complex treasury ops"
})
3 ways to find the same row. Keyword for exact, filter for attribute, semantic for conceptual. All pushed to SQL + pgvector.
02 · Company intelligence

24,000+ companies scored for agent readiness, free to query.

We scan every company's homepage, llms.txt, mcp.json, OpenAPI spec, and structured data signals — then blend them into a 0–100 AR score. Open layer, MCP-native, updated daily. Get context on a company before your agent loads their tab.

  • company_brief — AR score, AI-native classification, ecosystem role
  • scan_agent_readiness — scan any URL on demand, get the report
  • search_businesses — filter by industry, location, AR score range
  • semantic_search — "show me AI-native fintechs with llms.txt"
Browse the Agent Ready 100 →
MCP call · free tier
// One call before any recruiter / sales / partner conversation
Use OnlyData: company_brief({
  domain: "airbnb.com"
})

// → returns:
{
  name: "Airbnb",
  agent_readiness_score: 47,
  ai_native: "integrated",
  agent_ecosystem_role: "consumer",
  super_category: "Travel & Hospitality",
  team_members: [/* ... */],
  signals: {
    llms_txt:   false,
    mcp_json:   false,
    openapi:    true,
    structured: true
  }
}
→ Every row carries the evidence behind the score. Recruiters + sales + agents act on signals, not hunches.
03 · People + recruiters

The recruiter layer built on verifiable AI signal.

Vetted active AI researchers, engineers, founders — seeded from arXiv bylines, top OSS contributors, and opt-in program rosters (MATS, Apart). Every row scored on 5 dimensions with source-aware archetype floors. Filter by transition state (just-left, founding, building). Post a job, get ranked candidates with contact paths.

  • post_job_opening + match_candidates_for_job — ranked in one call
  • ai_minds_at_company — "who's at Anthropic, who's leaving, who's founding"
  • query_custom_dataset with transition_status + contactable filters
  • Person AR score with external_score (80) + platform_score (20) — earned signal, not claim-gated
See the recruiter flow →
MCP call · the money shot
// Recruiter flow: post a role, get ranked candidates
Use OnlyData: post_job_opening({
  company_name: "Contoso AI",
  title: "Founding Research Engineer — Interp",
  description: "mech interp + SAE + evals",
  role_type: "research_engineer",
  compensation_max_usd: 320000
})
// → then:
Use OnlyData: match_candidates_for_job({
  job_id: "...",
  only_transitioning: true,
  require_contact:    true,
  limit: 10
})
→ Live result: John Yang (Stanford, ex-Princeton · recently-left), Mikio Braun (ex-Zalando · recently-left), Buck Shlegeris (founding Redwood · founding). Each with contact path + AR + boost rationale.
04 · Applied consulting

Work directly with Cam + Jody on an agent-first rebuild.

Two paid slots a year at $20k/mo. We embed with your team — treat your company the way we treat OnlyData: MCP-native, agent-first, every surface callable by an LLM. One slot is live with Sturdy (BuildOps-adjacent ecosystem work). The other is open. Plus two mentorship tiers (unpaid, Boise-first) with limited spots.

  • Consulting · $20k/mo — 1 of 2 paid slots currently open · live-checked
  • Cam · mentorship2/2 spots taken · waitlist open
  • Jody · mentorship0/2 spots taken · apply below
  • Same intake form for all four tracks — we triage every submission by hand
Apply → See Sturdy example →
what one slot looks like
// Week 1 — audit
scan_agent_readiness({ domain: "yourco.com" })
company_brief({ domain: "yourco.com" })
// → score, gaps, adjacent ecosystem

// Weeks 2-4 — ship the spine
create_custom_dataset({ csv: "your first-party data" })
// → MCP-queryable, webhooked, embeddable

// Month 2+ — agent surface + team enablement
// your CEO and PMs call your own data
// from Claude, Cursor, Raycast — no new UI.
→ We treat your stack the way we treat ours. Outcome: your team ships MCP tools, your agents stop scraping, and your first-party data is the moat.

Three flavors, one MCP, zero new tabs.

Every surface above — custom datasets, company AR, person AR, recruiter tooling — is an MCP call. Your Claude / Cursor / Raycast / Rally / Ampcode agents already know how to reach it. No new dashboards, no new logins, no scrape-and-paste pipelines. Your agent writes the query, OnlyData returns the answer, the evidence stays public.