Connect to AI agents
Wire AdCrunch into Claude, Cursor, and ChatGPT via the MCP server. Tool reference and recipes.
Auth & scopes
- Auth & scopes
How OAuth works for the MCP server, scopes granted, token lifecycle, revocation.
Overview
- What is MCP
How the Model Context Protocol lets AI agents query AdCrunch in natural language.
Recipes
- Top campaigns yesterday
A one-line prompt to surface the best-spending campaigns of the previous day. - Meta vs TikTok ROAS
Side-by-side ROAS comparison across two providers for the same window. - Drill into one ad group
Find the underperforming creative inside a campaign. - Save and run a Skill
Teach the agent a workflow once, save it as a Skill, then run it by name on any future session.
Set up your AI client
- Claude Desktop
Install the AdCrunch MCP server in Claude Desktop and authorize your organization. - Claude Code
Install the AdCrunch MCP server in Claude Code and authorize your organization. - Cursor
Install the AdCrunch MCP server in Cursor and authorize your organization. - ChatGPT
Install AdCrunch as a custom MCP connector in ChatGPT.
Tool reference
- list_advertisers
List the advertisers (ad accounts) in your organization, optionally filtered by provider. - list_entities
List ad entities (campaigns, ad sets, ad groups, ads, creatives, …) under one advertiser, filtered by type or parent. - get_entity
Fetch the full raw provider payload for one ad entity by provider, type, and id. - query_insights
Aggregate spend / impressions / clicks / conversions over a date range, by entity. - skill_list
List your organization's ad-ops playbooks (Skills) by slug, name, and description. - skill_get
Fetch one ad-ops playbook (Skill) by slug, including its full instructions and current revision. - skill_create
Create a new ad-ops playbook (Skill) in your organization from an agent session. - skill_update
Update an ad-ops playbook (Skill), guarded by base_revision so concurrent edits aren't clobbered. - skill_delete
Delete an ad-ops playbook (Skill), guarded by base_revision.