Connectors

Agents: the full documentation index is at /llms.txt.

Revcenter runs a hosted MCP server. Add it to your AI app as a custom connector and the app can draft criteria, run searches and read shortlists in your workspace, with nothing to install.

https://api.revcenter.ai/mcp

When you add it, the app sends you to Revcenter to sign in. Pick the workspace to connect and click Authorize. Any member, Admin or Editor, connects apps, and the workspace needs a plan or credits. The app acts as you, with your role: an Editor’s connection sees only their own projects.

Add it to your app

Claude
ChatGPT

Open Customize → Connectors, click + Add and choose Add custom connector. On Team and Enterprise, an owner adds it first under Organization settings → Connectors (Add → Custom → Web), then members click Connect.
Name it Revcenter, paste the URL and click Continue. Keep the detected sign-in settings and click Add.
Click Connect, sign in to Revcenter and pick the workspace.

Works in Claude on the web, desktop and mobile. On the Free plan you can add one custom connector.

Any other MCP client that speaks Streamable HTTP and OAuth can connect the same way. A client that can’t sign in through a browser uses an API key instead: see Custom LLMs and agents.

Custom LLMs and agents

Model APIs, IDEs and agent frameworks can’t open a browser to sign in. Give them the same URL and a Revcenter API key as a bearer token:

URL:     https://api.revcenter.ai/mcp
Header:  Authorization: Bearer rvc_live_…

Get an API key

Any member creates one under Settings → Integrations → API (Authentication); it acts as them, with their role. A key made there holds every agent scope (projects, searches, shortlists and webhooks, each :read and :write), which is what the tools need. A key with fewer scopes lists only the tools its scopes reach, and a call beyond them is refused with insufficient_scope, exactly as the API refuses it. The key works at /mcp as it does at /v1: the same plan rule (plan_required without a plan or credits), rate limits, expiry and revocation, and its tool calls appear in the workspace’s API logs under the key’s name. Send it only in the Authorization header; a key in the URL is never read. Keep it in an environment variable or secret store, never in code or a prompt.

OpenAI Responses API

The mcp tool sends authorization to the server as the bearer token.

curl https://api.openai.com/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "model": "gpt-6-astra",
    "tools": [{
      "type": "mcp",
      "server_label": "revcenter",
      "server_url": "https://api.revcenter.ai/mcp",
      "authorization": "'"$REVCENTER_API_KEY"'",
      "require_approval": "never"
    }],
    "input": "List my Revcenter projects."
  }'

import os
from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-6-astra",
    tools=[{
        "type": "mcp",
        "server_label": "revcenter",
        "server_url": "https://api.revcenter.ai/mcp",
        "authorization": os.environ["REVCENTER_API_KEY"],
        "require_approval": "never",
    }],
    input="List my Revcenter projects.",
)
print(response.output_text)

Anthropic Messages API

The MCP connector is in beta: send the anthropic-beta: mcp-client-2025-11-20 header (or betas in the SDK), name the server in mcp_servers with the key as authorization_token, and enable its tools with an mcp_toolset entry.

curl https://api.anthropic.com/v1/messages \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "anthropic-beta: mcp-client-2025-11-20" \
  -d '{
    "model": "claude-opus-5-5",
    "max_tokens": 2000,
    "messages": [{"role": "user", "content": "List my Revcenter projects."}],
    "mcp_servers": [{
      "type": "url",
      "url": "https://api.revcenter.ai/mcp",
      "name": "revcenter",
      "authorization_token": "'"$REVCENTER_API_KEY"'"
    }],
    "tools": [{"type": "mcp_toolset", "mcp_server_name": "revcenter"}]
  }'

import os
import anthropic

client = anthropic.Anthropic()
response = client.beta.messages.create(
    model="claude-opus-5-5",
    max_tokens=2000,
    messages=[{"role": "user", "content": "List my Revcenter projects."}],
    mcp_servers=[{
        "type": "url",
        "url": "https://api.revcenter.ai/mcp",
        "name": "revcenter",
        "authorization_token": os.environ["REVCENTER_API_KEY"],
    }],
    tools=[{"type": "mcp_toolset", "mcp_server_name": "revcenter"}],
    betas=["mcp-client-2025-11-20"],
)
print(response.content)

Cursor and VS Code

{
  "mcpServers": {
    "revcenter": {
      "url": "https://api.revcenter.ai/mcp",
      "headers": { "Authorization": "Bearer ${env:REVCENTER_API_KEY}" }
    }
  }
}

{
  "inputs": [
    { "type": "promptString", "id": "revcenter-key", "description": "Revcenter API key", "password": true }
  ],
  "servers": {
    "revcenter": {
      "type": "http",
      "url": "https://api.revcenter.ai/mcp",
      "headers": { "Authorization": "Bearer ${input:revcenter-key}" }
    }
  }
}

Cursor reads ${env:REVCENTER_API_KEY} from your shell environment; VS Code asks for the key once and stores it.

Any other MCP client

Windsurf, LangChain’s MCP adapters and any other client that speaks MCP over Streamable HTTP connect the same way: the URL, transport streamable_http (or http), and the header Authorization: Bearer <key>. The server is stateless: POST JSON-RPC and each request is answered on its own, as JSON; there is no session id and no GET stream. Paid actions still ask first: a launch returns pending_approval with a proposal and runs only when your agent calls apply_proposal, so have it show the proposal to a person before it does.

What it can do

The same tools as the local MCP server, with two differences:

No file exports. A chat can’t save files, so the export tools are left out. Page through people with list_people, list_search_records and list_candidates, and send a client get_share_link.
Spending asks first. launch_search, launch_searches and expand_search reserve credits; publish_shortlist and send_shortlist_email reach the client; delete_project can’t be undone. Each one runs nothing at first: it returns pending_approval with a proposal and what to show you. The app runs it with apply_proposal only after you agree. Repeating a launch returns the same search and never pays twice.

Searches take minutes. Ask the app to wait for the search (wait_for_search waits up to three minutes a call), then to list everyone scoring 80 or more.

Disconnect

Settings → Integrations → Connectors lists the connected apps. Revoke stops one at once. Each connection is also an API key in Settings → Integrations → API, named after the app; revoking it there does the same. Connecting the same app again replaces its earlier connection.

Security

Sign-in uses OAuth 2.1 with PKCE. The app never sees your password or an API key. Custom LLMs and agents use an API key you give them instead; revoke it in Settings → Integrations → API to cut them off.
An access token lasts an hour and works only at https://api.revcenter.ai/mcp, for the one workspace you picked. The app renews it with a refresh token that changes on every use; a refresh token used twice disconnects the app.
Every call runs through the Revcenter API as that connection: the same permissions, plan, credits and rate limits as an API key.

Get started

Concepts

Tools

Guides

Recipes

Use cases

API reference

Add it to your app

Custom LLMs and agents

Get an API key

OpenAI Responses API

Anthropic Messages API

Cursor and VS Code

Any other MCP client

What it can do

Disconnect

Security

​Add it to your app

​Custom LLMs and agents

​Get an API key

​OpenAI Responses API

​Anthropic Messages API

​Cursor and VS Code

​Any other MCP client

​What it can do

​Disconnect

​Security

Add it to your app

Custom LLMs and agents

Get an API key

OpenAI Responses API

Anthropic Messages API

Cursor and VS Code

Any other MCP client

What it can do

Disconnect

Security