Connect Agent Planners to Claude: run Google Ads and GA4 from your chat
Claude speaks MCP and Agent Planners is an MCP server, so the two connect directly. Add the endpoint as a custom connector and sign in — no key to copy — and Claude can call the platform's ad and analytics tools — or hand it an entire goal and poll the result. Your ad-platform credentials stay on the platform; Claude never sees them. Writes follow the key's write mode — queued for an admin's approval, or run inside the organization's autonomy guardrails — and a write a platform marks as a person's decision never applies without a person either way.
What you can ask for once it is connected
- "Audit my Google Ads account for wasted spend over the last 30 days."
- "Pull GA4 sessions and key events by channel for last month."
- "Which search terms should become negatives?"
- "Start a task to rebalance budget toward the best ROAS campaigns" — and approve the writes in Agent Planners.
How to connect
In Claude, open Customize → Connectors → Add custom connector, enter https://www.agentplanners.com/api/mcp and select Connect: Claude opens the Agent Planners consent page, where you sign in, choose the workspace and approve what Claude may do. Claude Desktop and the mobile apps use the same connectors as your claude.ai account.
In Claude Code, run claude mcp add --transport http agent-planners https://www.agentplanners.com/api/mcp, then /mcp to sign in. An API key sent as Authorization: Bearer still works where you prefer one; OAuth sign-in has the details.
Call describe_permissions first in any new conversation: it reports the key's mode, scopes, plan entitlement and connected accounts, so the assistant knows what it may do before it tries.
Security
- The connection carries scopes — the ones approved on the consent page, or the key's — and without a write scope Claude cannot write no matter what is asked.
- Ad-platform credentials are never returned to the client — the platform injects them.
- Writes run inside the organization's guardrails and are audited with the same trail as the app.
Frequently asked questions
- Do I need the Claude connector directory?
- No. A custom connector works today with the endpoint URL alone — Claude signs in with OAuth. Agent Planners is not listed in the Claude directory yet.
- Does Claude see my Google or Meta credentials?
- No. They stay on the platform and are injected server-side; the client receives results, never secrets.
- Can Claude change my campaigns without asking?
- Only if the connection allows it. Without tools:write Claude cannot change anything. A connection approved on the consent page queues each change for an admin or editor by default, unless a workspace admin chose to apply ordinary changes at once. An API key's writes follow the mode it was minted with — queued for an admin, or run inside the organization's autonomy guardrails (magnitude caps, protected names, the monthly spend cap). A write a platform marks as a person's decision, such as a permanent delete, never applies without a person either way, and every write is audited.
- Which plan do I need?
- Agent mode (hand over a goal) works on any plan, with OAuth sign-in or an API key. Direct tool calls (tools mode) require a paid plan.