Model Context Protocol
Inrick, inside the assistant you already use
MCP is the open standard for giving an assistant real tools. Connect Inrick once and Claude, ChatGPT, Cursor or your own agent can find people, enrich companies and read your tables, using your workspace's credits and nobody else's.
https://inrick.com/api/mcp
What the assistant can do
Nine tools, the same ones the REST API exposes, with the same limits and the same audit trail. A tool that spends credits says so before it runs.
search_people
Find people
Find decision makers by title, seniority, department, company size, industry and location. Counting matches is free; returning rows uses credits.
Spends credits
search_companies
Find companies
Find companies by industry, headcount, location, funding and technology. Counting is free; returning rows uses credits.
Spends credits
find_work_email
Find a work email
A verified work email from a LinkedIn profile. Charged only when one is found.
Spends credits
enrich_person
Enrich a person
Full profile and career history for a person. Charged only when a match is found.
Spends credits
enrich_company
Enrich a company
Firmographics for a company from its domain or LinkedIn page. Charged only when a match is found.
Spends credits
list_tables
List tables
Every table in the workspace, with row counts. Free.
Free to call
read_table_rows
Read a table
Rows from one table, a page at a time. Free.
Free to call
get_account
Credit balance and plan
The workspace's plan, credit balance and period. Free.
Free to call
get_usage
Where credits went
What was spent, on what, over a period. Free.
Free to call
Set it up in a few minutes
Pick your client. The steps are the same ones you will find in Settings, API, where your own key and server address are filled in for you.
Add the server
claude mcp add --transport http inrick https://inrick.com/api/mcp --header "Authorization: Bearer YOUR_API_KEY"Check it connected
claude mcp listAsk for something
Try "find VP of Sales at Series B fintech companies in the US".
You need an API key with the scopes you want the assistant to have. Create one in Settings, API, then paste it in place of YOUR_API_KEY.
An assistant cannot spend what you have not allowed
A key carries only the scopes you give it, and a daily credit cap you set. Past the cap the key is turned away rather than charged, and it starts again at midnight UTC. Every call is in the request log with the key that made it.
Scopes, not blanket access
A key for search cannot enrich, and a key that can read tables cannot write to them.
A daily spending cap
Set what a key may spend in a day. An assistant deciding to enrich a thousand rows is one sentence away, and the cap is what stands between that sentence and your balance.
Revocable in one click
Revoke a key and every assistant using it stops at once, without touching the others.
Questions
- Which clients work?
- Any MCP client that speaks HTTP. Claude Code, Claude Desktop, ChatGPT, Cursor, VS Code and Windsurf have their own steps below; anything else uses the generic configuration.
- Does it cost extra?
- No. MCP calls spend the same credits as the same work done in the app or over the REST API, at the same rates. Counting matches is free.
- Can the assistant see my whole workspace?
- Only what the key's scopes allow. Tables are readable only with the tables scope, and nothing an assistant does is hidden from the request log.
- What happens when credits run out?
- The tool answers with the shortfall rather than failing silently, so the assistant can tell you instead of quietly skipping rows.
7 clients with their own steps above.