ServiceGraph
ServiceGraph · AI agentsVol. I  ·  No. 02

Stop scraping. Your agent's directory of US service firms is here.

ServiceGraph is a structured catalog of 118,000+ US professional-services firms — law, marketing, consulting, accounting, IT, architecture, engineering, HR, PR, design — pluggable into any AI agent.

$ npx skills add nostrband/servicegraph

Free tier, no credit card required.

Available for 19 agents

I.

What it looks like

One prompt. One filter. Ranked picks.

agent-session.log

user> Find me three boutique IP law firms in California that handle patent prosecution for hardware startups.

[skill: find-service-providers]

GET /v1/explore?filter=industry:legal+state:CA+service_provided:patent-law
→ 37 firms in pool
GET /v1/search ... & /v1/get for top picks
assistant> 3 picks for you:
  • Knobbe Martens— knobbe.com — (949) 760-0404  ★★★★★ 4.7 (Clutch · 142)
  • Carr & Ferrell— carrferrell.com — (650) 812-3400  ★★★★ 4.5 (G2 · 31)
  • Hickman Becker Bingham— hickmanlaw.com — (650) 314-0700  ★★★★ 4.6 (Clutch + Avvo · 87)
II.

Who it's for

Three integrations, one catalog.

Procurement agents

Help users find and shortlist vendors with specific industry, service, location, size, and rating criteria — without sending them to a Google search.

CRM enrichment

Pass a list of company domains, get back classified industry/services/contact in one structured response. Bring-your-own-list works without /search.

Sales prospecting

Build target lists in seconds: every IT consultancy in TX with at least one US-based engineer team and a third-party rating.

III.

What you get for each firm

Real structured data — not search snippets.

Identity & classification
  • ·Industry across 22 categories (legal, marketing, consulting, IT, …)
  • ·Multi-tag service taxonomy (89 tags) with per-tag evidence level — low / medium / high
  • ·Service model — multi-person firm, solo consultant, agency, nonprofit
  • ·Geography served — city / multi-state / national US / international
  • ·Company size signal — solo / 2–10 / 10–50 / 50+
Contact & outreach
  • ·Primary phone (US-NANP filtered) and email
  • ·Full address — street, city, state, ZIP — where disclosed
  • ·Website URL and brand-name + legal-name (where extractable)
  • ·Social presence — LinkedIn company page, Twitter/X, Facebook, YouTube, GitHub, TikTok, …
Reviews & reputation
  • ·Aggregated client rating + total review count, averaged across third-party directories
  • ·Presence flags for individual directories — Clutch, G2, Capterra, Yelp, BBB, GoodFirms, TheManifest, Trustpilot, …
  • ·Employer ratings — Glassdoor and Indeed, with review counts
  • ·Wikipedia / Crunchbase / LinkedIn-company presence signals
Engagement & pricing
  • ·LinkedIn followers + employee count where reported
  • ·Facebook engagement (likes, talking-about) where the page exists
  • ·Founded year and other vintage signals
  • ·Pricing model + range where the firm publishes it (hourly, project, tiered, quote-on-request)
IV.

Sample prompts

Click any prompt to see the run-through.

FAQ

Honest answers.

Do I need an API key to try it?
Yes. Every call from a skill, MCP server, or your own script authenticates with an API key. Create one at /profile/api-keys after signing in. The key is what your agent sends as a bearer token.
What does the free tier cover?
Every account gets free credits on signup. 1 credit = 1 full company profile (contact, team, case studies, sources). Search and browse on the website are always free; only API responses from /v1/get consume credits.
How is firm data sourced and how fresh is it?
The catalog is built from public web data — firm websites, third-party directories, public business registries — and refreshed continuously. Classification (industry / services / size / geography) is done with LLMs over the firm's own pages and is re-checked on update. Stale rows are dropped; new firms enter as they're crawled.
Which AI agents work?
Anything that supports the Skills convention — Claude Code, Codex, Cursor, GitHub Copilot, Gemini, Cline, Droid, Goose, Windsurf, and a dozen more. The skill is plain markdown + curl invocations; agents that don't support skills can call the API directly.
Is the data US-only?
Yes for v1. The catalog explicitly excludes non-US firms; it's scoped to US professional-services as a focused, defensible target. Other regions are on the roadmap.
How is this different from a Google search?
Structured filters, not keyword matching. Industry across 22 categories, services across an 89-tag taxonomy with evidence levels, classified firm size, third-party rating presence, geography served — all queryable with a single filter string. You ask for “mid-size IT consultancies in TX with a Clutch listing and 4★+ rating” and get a clean list, not 200 search results to re-rank.
What happens if a firm I want isn't in the catalog?
/v1/get returns a 404 (uncharged) so your agent can branch. Coverage is ~118k+ US firms today and growing; if you need a specific firm indexed, email artur@servicegraph.co and we can prioritize.