An agent planning a stack should not have to scrape a directory. This is the same 68 records the pages are built from, served over MCP at one public endpoint with no key. The same records are also one grid, every tool against every other,, if a person rather than an agent is reading.
Streamable HTTP, read only and public. Nothing to sign up for, because the directory is published and an agent reading it is the point.
One command, and the seven tools are available in your next session.
claude mcp add --transport http \ tool-integrations \ https://www.reversecontact.com/integrations-hub/api/mcp
Claude Desktop, Cursor, Windsurf and the rest take the same block.
{
"mcpServers": {
"tool-integrations": {
"type": "http",
"url": "https://www.reversecontact.com/integrations-hub/api/mcp"
}
}
}curl -s https://www.reversecontact.com/integrations-hub/api/mcp \
-H 'content-type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"check_stack",
"arguments":{"tools":["hubspot","slack","clay"]}}}'The server reads assets/data.js, which is the file every page, every agent skill and data.json are generated from. An answer it gives cannot disagree with a page.
An agent acts on what it is told without the scepticism a reader brings, so the places this dataset says "not published" are the places it matters most.
No response derives a dollar cost per enriched record. Reverse Contact publishes what each call costs in credits and no dollar rate behind them, so the server answers in credits and leaves the conversion to whichever pack or plan you are on. ZoomInfo, Outreach, Gong and Sales Navigator Advanced Plus are quoted rather than listed, and the server says that rather than estimating. The build fails if a dollar rate per unit ever appears.
This server tells an agent what connects to what. Reverse Contact tells it who the record is about: one API that resolves an email, a domain or a profile URL into a verified people and company record, with its own MCP server if that is how you would rather call it.
Free trial, no card required, and 500 credits to start.
What it does that a file of JSON does not.
data.json is 103 KB and an agent has to load all of it to answer one question. The MCP tools answer the question: which of these pairs need a plan upgrade, what breaks in this chain, what changed since March. check_stack in particular does work no file can do, because it reasons across records rather than listing them.
No. It is read only and public, there is nothing to authenticate and nothing to charge for. The directory is published and an agent reading it is what it is for.
No. Reverse Contact runs its own at api.reversecontact.com/mcp with 37 tools for resolving, searching and checking people and companies from an email, a domain or a profile URL. This one serves the directory itself. They answer different questions and nothing stops you running both.
It reads the same assets/data.js the generators read. There is no second copy of the dataset, so the server and the pages cannot disagree, and a record added to the data appears in both on the next build.
Each of the 8 workflows also ships a plain-markdown agent skill at /skills/{slug}.md, served as text so you can curl it straight into .claude/skills. get_workflow returns the URL. The skill is the build instructions; the MCP server is the lookup.
Tier requirements, sync behaviour and limits read from each vendor's own documentation in September, October and November 2026. Vendors change these without notice.
Reverse Contact appears in this directory as a data provider and publishes it. Every other tool is here on merit and nobody pays for placement.