Connecting RB2B's Profiles API to Claude
Install one small skill file and Claude can fetch your RB2B visitor profiles directly — paging through results, respecting rate limits, and keeping your API key out of the conversation.
Download skillWhat it does
This skill gives Claude a working connection to RB2B's Profiles API. Once it's installed, you can ask Claude for your identified visitors in plain English — “who from fintech visited this week”, “export last month's VPs to a CSV”, “summarise which companies keep coming back” — and Claude will fetch the profiles, page through them, and work on the results for you.
It's deliberately small. The whole skill is a single SKILL.md file that tells Claude where the API is, how to authenticate, how paging and rate limits work, and how to handle the results without flooding the conversation.
What you can ask it
Anything that starts with “my RB2B visitors”. A few that work well:
- Summaries. “Summarise my RB2B visitors from the last 7 days by industry and seniority.”
- Filters. “Which directors and above from companies over 500 employees visited since Monday?”
- Exports. “Pull every profile from September and save it as a CSV.”
- Patterns. “Which companies had more than one person visit this month?”
The API itself only filters by date, so Claude fetches the date range you ask for and applies everything else — industry, title, company size — after the profiles are downloaded. That's built into the skill, so you don't need to phrase requests any particular way.
Installing it
Download the skill using the button on this page. You'll get a zip containing one folder, rb2b-profiles, with the SKILL.md file inside.
Claude.ai and Claude Desktop. Open Settings → Capabilities, make sure Skills are enabled, then choose Upload skill and select the zip. Claude also needs to be able to make web requests and run code, so leave code execution switched on.
Claude Code. Unzip the download into ~/.claude/skills/ for every project, or into .claude/skills/ inside a single repository. Claude Code picks it up on the next session.
Your API key
The skill needs an RB2B API key, which you'll find in your RB2B account settings. API access depends on your plan — if your account doesn't include it, the API returns 403 api_access_not_allowed and Claude will tell you that rather than retrying.
In Claude Code, the cleanest option is to set the key as an environment variable before starting a session:
export RB2B_API_KEY=your-key-here
If the variable isn't set, Claude will ask for the key in the conversation instead. Either way, the skill instructs Claude never to repeat the key back or write it into a file.
How it handles paging and limits
The Profiles API returns up to 100 profiles per request, newest first, with a cursor pointing at the next page. The skill tells Claude to always ask for the maximum, follow each cursor until there are no more pages, and never construct a cursor itself — which is the most common way an API client quietly skips or repeats records.
The API allows 20 requests a minute. If Claude hits the limit, it waits the number of seconds the API asks for and resumes from the same cursor, so large pulls finish rather than failing halfway through.
There's one hard boundary worth knowing: the since date can't be more than 90 days back. For anything older, use an RB2B export rather than the API.
Keeping results out of the chat
A month of visitors can run to thousands of profiles. Pasted into a conversation, that burns through Claude's context and makes the answer worse, not better. The skill tells Claude to save results to a file and analyse them with code, and only show you raw profiles when you ask for them.
In practice that means you get the answer — a summary, a ranked list, a CSV — rather than a wall of JSON you then have to ask Claude to make sense of.
Before you run it
Start with a narrow request, such as yesterday's visitors, to confirm the key works and to see the shape of the data your traffic produces. Once that comes back cleanly, widen the date range.
If your results look thinner than expected, the problem is usually upstream of the API. Restricting collection by URL and excluding internal domains shape what RB2B collects in the first place, and therefore what this skill can return.