Cofoundr MCP

Connect your AI to your advertising operation.

82 Meta actions available through a single MCP endpoint. Read performance data, manage campaigns, control budgets, build audiences. All from Claude, ChatGPT, or any MCP-compatible AI.

// "Find campaigns where CPL spiked"
❯ tool: get_campaigns_with_adset_results
❯ account: "act_123456789"
// 3 campaigns with CPL > 30% above target
// "Pause the worst ad set"
❯ tool: pause_object
❯ id: "23851234567890"
✓ Paused. Logged to audit trail.
// "Increase budget on the winner by 20%"
❯ tool: update_budget
✓ Budget updated. ₹500 → ₹600/day

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI tools connect to external data and services. Cofoundr runs an MCP server that gives your AI direct, secure access to your Meta ad accounts. Your AI assistant can now read your ad data and take action on it. No CSV exports, no copy-pasting, no middleware.

82 actions. One endpoint.

Everything your AI needs to read, manage, and act on your Meta ad accounts. Organised by what they do.

Read data 22 actions
  • Campaign, ad set, and ad performance
  • Spend, results, ROAS, CPA, CTR
  • Health scores and diagnostics
  • Wasted spend detection
  • Period comparisons and trends
Campaign management 18 actions
  • Create, update, duplicate campaigns
  • Pause and activate ad sets and ads
  • Update targeting and placements
  • Manage campaign objectives
  • Batch operations across accounts
Audiences 12 actions
  • Create custom audiences
  • Build lookalike audiences
  • Estimate audience size
  • Manage audience users
  • Get audience insights
Creative and media 14 actions
  • Upload images and videos
  • Create ad creatives
  • Get ad previews
  • Creative performance diagnostics
  • Hook rate, hold rate, CTR per ad
Budget and rules 9 actions
  • Update daily and lifetime budgets
  • Create ad schedules
  • Manage automation rules
  • Set bid strategies
  • Burn rate monitoring
Reporting and insights 7 actions
  • Dashboard and overview data
  • Period comparison reports
  • Wasted spend summaries
  • Page and Instagram insights
  • Account-level analytics

What you can ask your AI to do

Real prompts. Real actions. This is what working with Cofoundr MCP looks like.

▶ "Find all campaigns where CPL increased more than 30%."
▶ "Pause the three worst-performing ad sets."
▶ "Increase the budget of the best campaign by 20%."
▶ "Create a lookalike audience from top converters."
▶ "Show me everything that changed yesterday."
▶ "Upload these creatives and create ads for them."
▶ "Compare this week to last week across all accounts."
▶ "Which accounts are wasting money right now?"

Live in under 2 minutes

Three steps to give your AI access to your Meta ad accounts.

1

Connect your Meta accounts

Log in with Facebook and select the ad accounts your AI should access. Same secure OAuth used by the Cofoundr dashboard.

2

Generate an MCP token

Create a token in Settings. Scope it to specific ad accounts. Control read and write permissions. Revoke anytime.

3

Point your AI at the endpoint

Add the MCP endpoint to Claude, ChatGPT, Cursor, or any MCP-compatible tool. Your AI can now read and act on your ad data. Step-by-step: Claude, ChatGPT, Cursor.

Built for trust

Your ad accounts contain real money and real client data. Security is not optional.

Per-token account scoping

Each token is scoped to specific ad accounts. Your AI only sees what you allow.

Every action logged

Full audit trail with timestamps, action type, and account. Know exactly what your AI did and when.

Revoke access anytime

Delete a token and access stops immediately. No waiting, no grace period.

Encrypted tokens

Access tokens encrypted at rest. Decrypted only in memory at the moment of a request.

Read vs write permissions

Control whether a token can only read data or also make changes. Start read-only, upgrade when ready.

Client data isolation

No client ever sees another client's data. Account boundaries are enforced at every layer.

Questions, answered

What is MCP?+
MCP (Model Context Protocol) is an open standard that lets AI tools connect to external data sources and services. Cofoundr runs an MCP server that gives your AI direct, secure access to your Meta ad accounts through a single HTTP endpoint.
Which AI tools work with it?+
Any tool that supports the Model Context Protocol. Claude Desktop, ChatGPT, Cursor, Windsurf, and any custom integration. You bring your own AI. We provide the data and action layer.
What are read vs write actions?+
Read actions pull data from your accounts: campaign metrics, health scores, diagnostics, reports. They are always free and unlimited. Write actions change something: pausing an ad set, adjusting a budget, creating a campaign. Each write action costs 1 credit.
Is it safe to let AI manage my ads?+
Every action is scoped, logged, and auditable. Tokens are per-account with revocable access. You control read vs write permissions. Your AI cannot access accounts outside the token scope or take actions you have not authorised.
Do I need the full Cofoundr dashboard?+
No. MCP is included in every Cofoundr plan, so you can use the endpoint on its own or alongside the dashboard. Connect Meta to start with 50 free credits for 7 days, no card.
How many actions are available?+
82 actions across 7 categories: read data, campaign management, audiences, creative and media, budget and rules, reporting, and insights. Every plan gets access to all 82 actions.

Your AI is ready. Your ad accounts are waiting.

Connect Meta, generate a token, and start managing ads from your AI in under 2 minutes. 82 actions. One endpoint.

Connect Meta. See what Cofoundr finds.
50 free credits for 7 days · MCP included in every plan · No credit card