>_

SpendSignoff documentation

The always-on operator for your ad accounts. Read and draft from your AI client — approve before anything goes live.

Agent setup

Run the agent in Claude Code

One CLI command registers the SpendSignoff MCP server in Claude Code. Authorize an ad account read-only from the SpendSignoff dashboard, then type plain-English instructions and approve drafts in the app.

Your AI can read and draft — it can never spend without your approval.

Claude Code is supported today

Claude Code is the terminal-based coding agent. It supports remote MCP servers via the claude mcp add command, which registers the server URL and runs the OAuth 2.1 handshake in one step.

Once registered, the SpendSignoff tools appear in every Claude Code session. The agent reads your linked Google Ads and Meta accounts and drafts changes; approval stays in the SpendSignoff app.

Register the MCP server

Run this command once from any terminal. Claude Code stores the server in your user config.

terminal

claude mcp add --transport http spendsignoff https://mcp.spendsignoff.com/mcp

First-time setup

1

Register the server

Run claude mcp add --transport http spendsignoff https://mcp.spendsignoff.com/mcp. Claude Code opens a browser tab for the OAuth 2.1 flow and requests mcp.read and mcp.draft scopes. There is no mcp.approve scope to grant.

2

Authorize an ad account

Open the SpendSignoff dashboard, go to Accounts, and connect a Google Ads or Meta account. The first link is read-only: the agent reads spend, structure, and performance data. Reads are ungated on every plan including Free.

3

Type a plain-English instruction

In any Claude Code session, ask the agent something like "show me which campaigns are pacing over budget this week." The agent calls get_account_snapshot and query_entities and returns a plain-English summary.

4

Review and approve your first draft

Ask the agent to propose a change — for example, "reduce the daily budget on the top-of-funnel campaign by 15%." The agent calls propose_change and returns a before-and-after diff. Open the Approval Queue in the SpendSignoff app, read the diff, and click Approve and push live, then Confirm.

Read and draft only — the agent cannot spend your budget

The SpendSignoff Performance Marketing Agent holds mcp.read and mcp.draft scopes only. It reads account data and produces change drafts. It cannot push a change live. Every live change requires a two-step Approve and Confirm in the SpendSignoff app and is recorded in a KMS-signed append-only audit log.

Verifying the connection

After registering, run claude mcp list to confirm spendsignoff appears in the server list with a status of connected.

To test without a live ad account, ask the agent "list my ad accounts." If the OAuth flow completed, it returns the accounts linked in your SpendSignoff dashboard. If it returns an auth error, re-run claude mcp add to refresh the token.

Verify the registration

Check the server is registered and connected:

terminal

claude mcp list

Next

Core workflows

The five operations — audit, launch, optimize, report, and monitor — with exact prompts.

    Run the SpendSignoff agent in Claude Code