The Permutive MCP server is in a testing phase with partnering customers. Access is invitation-only and the
tool interfaces may change. If you would like access, please contact your
Permutive representative.
What it is
The Permutive MCP server gives AI agents a direct, structured line into your audience data. Connect an agent — such as Claude — to your Permutive account, and it can explore and reason about your cohorts in plain language: find audiences that match a brief, pull their detail, and combine them to check reach, all without you writing queries or leaving the conversation. It implements the Model Context Protocol, an open standard — published by Anthropic and adopted across the major model providers and agent frameworks — for connecting AI agents to external systems. The shorthand is “USB-C for AI”: instead of a bespoke integration per agent, the agent and the service agree on a common protocol. The agent reads the tool list, understands each tool from its description, and calls it — no glue code required.An open protocol
MCP is vendor-neutral. The server works with Claude, ChatGPT, Cursor, and any
agent or framework that speaks the protocol — you are not locked into a model
provider to use it.
Structured, not scraped
Every tool has a typed schema and a description the agent reads at runtime.
No HTML parsing, no fragile screen-scraping — the agent gets a precise,
typed definition of each tool’s inputs and outputs.
Organization-scoped
Authentication is per organization. An agent acting on your behalf sees only
your data — your cohorts, domains, and activations. The same boundaries that
apply in the Permutive UI apply here.
What you can do
The tools cover the full discovery-to-evaluation loop a commercial team runs when responding to a brief:- Discover audiences — find cohorts matching a natural-language brief (“high-income readers interested in luxury cars”) across one or more workspaces.
- Understand a cohort — look up a cohort’s name, type, state, tags, and which workspaces it is available in.
- Measure reach — estimate how many users and page views an audience reaches, with optional breakdowns by cohort, domain, device, country, and workspace.
- Compose audiences — combine cohorts and measure the result, iterating as you refine against the brief.
Example use cases
- A planner sizing an audience for a campaign brief, directly in their AI assistant.
- An analyst comparing the reach of several cohort combinations without building a report.
- Quickly checking which workspaces a cohort is available in before activating it.