How it works
Your agent app talks to RevyOps through a secure MCP connection. Ask a question, and the agent pulls live workspace data to answer in plain English.What your agent is told when it connects
Every agent receives the same short briefing the moment it connects, whichever app it runs in:- Start with the Account Setup Guide in the Guides folder, and read long documents one section at a time.
- Write the workspace’s answers in Your Setup, never in Guides.
- Look things up in these docs, using the machine-readable index at
https://docs.revyops.com/llms.txt, before saying something is not possible. - Use the REST API when no tool fits. Creating, upserting and deleting contacts and companies, merging and splitting people, creating and deleting emails, and logging calls, emails, LinkedIn messages, SMS and WhatsApp are available there with the same key.
- Keep a filter as a saved view with
save_viewoncequery_entityreturns the right records, so the view shows in the app’s sidebar. - Build workflows with the workflow tools, on a saved view of this workspace, one step at a time, and ask you before turning one on or testing a Webhook or Alert step, because both really send.
- Never send a message to a person. The agent drafts, ranks, researches and prepares; you send.
Set up in three steps
1
Choose how your agent signs in
Agents that support sign-in, such as Claude Code, can connect with an Agency admin’s or Workspace admin’s RevyOps login: you approve the workspaces and access in your browser, and no key is needed. Otherwise, use an API key. Go to the Keys tab on the Developer page and click Generate key. Use a read-only (
RO-) key if the agent only needs to read, or a read-write (RV-) key if it should also make changes. An agency master (ML-) key doesn’t work here: it gives the agent no workspace. See Authentication.2
Install the agent app of your choice
For example, Claude Desktop.
3
Add the RevyOps MCP server to your agent
Point your agent at
https://mcp.revyops.com/mcp and send your key in the X-API-KEY header, then restart the app. Follow the guide for your agent:Claude Desktop
Claude Code
Other agents
Example questions to ask
- “Show me all campaigns from last month and their reply rates”
- “Which sending domains have the best deliverability right now?”
- “Find all contacts at acme.com and show their current status”
- “What’s our interested rate this week compared to last week?”
- “Drill into campaign [name] and show me everyone who replied”
- “Which contacts replied on LinkedIn this week?”
- “Find our ICP guide in AI Documents and summarize it”
- “Set up a workflow that posts to Slack when a contact is marked interested”
Query tips for best results
- Be specific with time ranges. “last 7 days” or “January 2026” works better than “recently”.
- Use campaign names. Reference campaigns by their exact name for precise results.
- Request comparisons. Ask your agent to compare metrics across time periods or campaigns.
- Export data. Ask your agent to format results as tables or CSV for easy sharing.
Available tools
The MCP server has 77 tools, grouped by area. See Tools for every tool and its parameters.Workspace access and key security
- Each workspace requires its own API key.
- You can use multiple keys at the same time in your agent app.
- When a question doesn’t name a workspace, the agent uses the workspace of the first key in the header. Name the workspace in your question, for example “campaigns for Acme last week”, whenever you have more than one key.
- Keep your API key private. Never share it publicly or commit it to version control.
- Regenerate keys regularly. Regenerate key issues a new secret and invalidates the old one.
- Revoke compromised keys immediately. Use Revoke key on the Keys tab, then generate a replacement.
Limits
Each API key can make up to 90 requests every 10 seconds across the API and the MCP server. See Rate Limits.Privacy and data access
- The connected model can only access data your API key has permission to view.
- Conversations are subject to your chosen agent provider’s privacy policy.
- No RevyOps data is stored permanently by the agent. Each session is independent.
- Your RevyOps data remains in your workspace and is not used to train AI models.

