MCP Server (Beta)Skills work with the hosted MCP server (recommended). They also work with a local server via npm or a source build; the npm package may lag behind hosted, so prefer the remote URL when your client supports it.
Available skills
Single Enrichment
Look up a person, company, email, or phone number using the best matching enrichment.
Table-Driven Enrichment
Create a table, insert rows, attach an enrichment, run it, and get a shareable link.
Waterfall Enrichment
Try multiple data providers in sequence to maximize success rate.
Bulk Enrichment
Enrich a list of up to 100 records in a single operation with inline results.
How skills work
Without skills, you have to guide the AI step-by-step: “search for an enrichment, then get details, then run it…” With skills, you just say “find the email for John Smith at Google” and the agent knows the full workflow automatically. Skills use a three-phase loading model to stay efficient:| Phase | What loads | Token cost |
|---|---|---|
| Discovery | Name and description only | ~100 tokens per skill |
| Activation | Full instructions when the task matches | Under 5,000 tokens |
| Execution | Agent follows the workflow | Zero additional overhead |
Skills are loaded on demand — they don’t consume context until the agent decides they’re relevant to your request.
Installing skills
Skills are included in theskills/ folder of the GitHub repo. To use them:
- Claude Code / Claude.ai
- Cursor / Codex / Other agents
- OpenClaw
Copy the skill folders into your Claude skills directory:Claude will automatically discover and activate the skills when relevant.