Many Whatagraph customers already use Claude, some with Whatagraph connected to it. So it's a fair question: if Claude can already reach your data, why add IQ Agents? The two solve different problems, and they work well together. If you're new to agents, start with What are IQ Agents?
The short answer
Claude is a personal assistant: one person works with it, session by session, bringing the instructions and context each time. An IQ Agent is closer to a team member. You describe the job once, it lives inside Whatagraph, where your whole team can use and improve it, and it runs on Whatagraph's servers, so the work happens whether or not anyone is at their desk. Connecting Claude to Whatagraph grants it access to your data, but you still provide the instructions, context, and supervision for every session. An IQ agent carries all of that with it.
Set up once, used by everyone
Each IQ agent is configured once with its own instructions, knowledge, and memory, like a job description. After that, a short request is enough, because the agent already knows its job. Everything it knows is shared, so anyone on the team can ask it and get the same result. And because the agent and everything it knows live in Whatagraph, the setup doesn't leave with the person who built it. Connecting tools works the same way. An admin connects a tool once through OAuth, and every teammate's IQ agents can use it, even people without their own login for that tool. For how this works, see How to connect external tools to your IQ Agents with MCP.
The numbers always match
IQ Agents use the same data your reports use, with the same definition of every metric: how channels are blended, which filters apply, what counts as a conversion, and your custom metrics. A general AI tool reading through a data connector gets the raw numbers without those decisions, so its answers can drift from what your client sees. Because agents inherit your data layer, their numbers are the report's numbers.
Control that your clients can trust
Every action an IQ agent takes is attributed to that agent. You decide, per agent, what it can do on its own, what needs your approval, and what's off-limits, and before it makes a change that affects your reports, it asks you to confirm. You can also save a report's state as a snapshot before an agent works on it. Every tool an IQ agent uses is a scoped permission that's logged and can be revoked, so an agent never holds your passwords.
IQ Agents take action, not just the reporting
Agents don't stop at reporting. They can run on a schedule with your laptop closed, then act on what they find: posting a summary to Slack, creating a task in a connected project tool, or taking action in another app you've connected. Because Whatagraph runs the connections and keeps your data fresh, your team spends its time on the work rather than on maintaining the setup.
You don't have to choose
IQ Agents aren't a replacement for Claude. Whatagraph connects to Claude via its MCP server, so both work with the same account. Use Claude for exploration and hand off the repetitive work to IQ agents. You can also build and manage your agents through Claude using the same connection.
Tip: Give an IQ agent its own knowledge about a client, and that context is shared with your whole team instead of living in one person's account.