Cursor, Claude Code & Windsurf

This article covers three ways to put that inside your coding agent: a paste-once prompt for Cursor, Claude Code, and Windsurf; the MCP server for Claude; and the REST API for scripts and terminals.

Pick your path

• Cursor, Claude Code, Windsurf — paste a prompt. The agent reads the docs, asks for a key, and runs a test. Start below.
• Claude (Desktop or Web) — add the MCP server as a custom connector.
• Scripts and terminals — call the REST API directly.

You will need an API key. Create one from your account at abm.dev, then keep it out of source control.

The fastest start: paste a prompt

Your agent can wire itself up. Paste this into Cursor, Claude Code, or Windsurf:

Set up abm.dev as an enrichment tool in this project.

1. Read https://abm.dev/llms.txt for the full agent context.
2. Tell me to create an API key, then paste it here.
3. Save the key in this project's .env (and add .env to .gitignore).
4. Run one test enrichment with my email.
5. Report what came back — and which source cited which field, with the confidence on each.

The agent reads the live context file, requests a key, saves it, runs a single enrichment, and reports back with citations. No copying schemas by hand. No guessing the request shape.

Claude: add the MCP server

In Claude, go to Settings → Connectors → Custom and paste:

https://mcp.abm.dev/mcp

The MCP server exposes abm.dev’s enrichment tools to Claude, so you can enrich a record inside a conversation. The exact tool list is unverified — see /for-agents for what is wired up.

Scripts and terminals: the REST API

Two endpoints. That is the whole surface you need.

Submit an enrichment:

curl https://api.abm.dev/api/v2/enrichments \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"targets":[{"type":"person","email":"[email protected]"}]}'

Each target also accepts a LinkedIn URL or a company domain. Use whichever you have.

Retrieve a result by id:

curl https://api.abm.dev/api/v2/enrichments/{id} \
  -H "x-api-key: YOUR_API_KEY"

What comes back

One record, resolved from ten sources — LinkedIn, Companies House, Perplexity, Tavily, Hunter, and others — into a single response. Eighty-nine canonical fields: forty-three for the person, forty-six for the company. Forty signals.

Every value carries three things:

• A source. Which provider produced it.
• A confidence, from 0 to 1. Higher means the sources agree; lower means they diverge or the evidence is thin.
• A selection_reason. Why that value was chosen over the alternatives.

A value is cited or it is not returned. No fabricated facts. No silent fallbacks. On the field set and how confidence is computed, see canonical fields, data sources, and confidence scores.

Agent loops and batch jobs

For live agent loops, abm.dev streams results in real time over SSE — your agent reads values as they resolve. For batch work, webhooks fire when a job finishes. The webhook contract — paths, headers, retries — is not documented here; see /api-reference/jobs for the current spec.

What it costs

Per enrichment. No subscription. Credits never expire, and all ten sources are included in every call — no per-source bills, no per-field charges. From about €0.29 per enrichment.

Packs:

The Playground is free — try a call before you spend a credit. New accounts get free launch credits with the code LAUNCHCODES. Full pricing lives at /#pricing.

Where to go next

• Getting started — the first call, end to end.
• API reference — the live contract. Start with enrichment.
• For agents — MCP, llms.txt, and the agent context.
• For developers — REST patterns and integration.
• CLI — the terminal client. (Command names and flags: see the page; specifics are not restated here.)
• Integrations — Claude connector, HubSpot, LinkedIn.

Built and operated by Foxley Farm Operations Ltd (UK Company No. 16392009), Oakham, United Kingdom. GDPR-compliant.

Stuck? Open a support ticket.