Servicegraph

What servicegraph MCP server does

The servicegraph MCP server gives an agent access to ServiceGraph datasets through one branded skill and several use-case-specific skills. Available data includes more than 110,000 US agencies, 3,500 business directories, 850 product directories, and 50,000 newsletters. Agency records include services, company size, location, and ratings. Directory records include measures such as domain rating and traffic, while product directories also include submission policy.

The branded servicegraph entry point is dataset-agnostic. It can explain which datasets are available, learn their schemas and filters from the API, search records, and retrieve additional details. Dataset-specific find-* skills currently cover the agencies and product-directories datasets.

How it works

An agent invokes the branded skill by referring to ServiceGraph, or uses a more focused skill such as find-service-providers, find-marketing-agency, or find-law-firm. The focused agency skills translate a request into filters, including industry, services, location, size, and ratings. Some skills automatically add relevant filters; for example, the SEO skill combines the marketing-agency industry with an SEO service filter.

Searches return brief rows without consuming credits according to the README. Contact details and metric-rich information require credits. The same API key and filter DSL apply across the datasets, so the generic skill can adapt as additional data becomes available.

Setup and configuration

Install one skill with:

npx skills add nostrband/servicegraph --skill find-service-providers

To install the complete collection, run:

npx skills add nostrband/servicegraph

Create an API key through the ServiceGraph profile page and set it as SERVICEGRAPH_API_KEY in the shell or an .env.local file. Every call requires a key, including browsing and filtering. New accounts receive 2,000 free credits without requiring a card. Skills request the key on first use and do not place its value in the model context.

Tools and capabilities

The servicegraph MCP server supports these practical workflows:

• Discover available datasets and their fields.
• Filter and rank US professional-services firms.
• Find agencies for marketing, SEO, design, software development, web development, AI consulting, legal work, accounting, and management consulting.
• Search product directories for software, SaaS, MCP-server, and AI-tool launches.
• Retrieve contact information for selected companies or domains.
• Enrich results with metrics such as domain rating, traffic, ratings, and company size where the dataset provides them.

The agency catalog is B2B-focused in areas such as law and accounting. Personal legal matters and individual tax preparation are outside the stated scope.

Limitations and notes

The README describes this repository as Agent Skills for ServiceGraph, so support depends on an agent harness that supports the Agent Skills format. Named compatible agents include Claude Code, Codex, Cursor, GitHub Copilot, Gemini, Cline, Goose, and Windsurf; only clients represented in the directory’s compatibility list should be selected for metadata.

The servicegraph MCP server does not provide anonymous access: an API key is needed for all calls. Detailed contact and metric data also consume credits. Subreddits and influencers are listed as planned datasets rather than currently available ones, and additional newsletter-related skills are described as forthcoming.