OrbiAds — Google Ad Manager MCP

A skill for Claude, ChatGPT, Gemini, and OpenAI Codex that gives your AI assistant direct access to Google Ad Manager.

→ Get started free at orbiads.com · ★ Star this repo

Install the skill

Claude Code (CLI) — one command:

./install.sh claude          # current project
./install.sh claude --global # all projects
./install.sh skills          # load all 8 skills for this session

Claude Desktop — add to claude_desktop_config.json:

{
  "mcpServers": {
    "orbiads": {
      "type": "http",
      "url": "https://orbiads.com/mcp"
    }
  }
}

claude.ai — Settings → Integrations → Add MCP server → https://orbiads.com/mcp

ChatGPT — Settings → Connectors → Create connector → MCP URL: https://orbiads.com/mcp

Gemini / AI Studio — Tools → MCP configuration → https://orbiads.com/mcp

OpenAI Codex — copy openai-codex/AGENTS.md and openai-codex/mcp/config.remote.json to your workspace root

→ Full guides: Claude · ChatGPT · Gemini · OpenAI Codex

What is OrbiAds?

OrbiAds is a hosted MCP server that connects your AI assistant directly to Google Ad Manager (GAM). Instead of clicking through the GAM interface or writing API scripts, you describe what you want in plain language — OrbiAds handles the API calls, guardrails, and audit trail.

You: "Check inventory availability on the homepage banner for a 300x250 in France next week"
OrbiAds: [runs forecast] → Available: 1.2M impressions. Pressure: low. Safe to traffic.

You: "Create the line item for Renault, €15 CPM, Monday to Friday"
OrbiAds: [applies guardrails] → Preview ready. Confirm to push?

No scripts. No API tokens to manage. No switching tabs.

Who is this for?

AdOps managers who traffic campaigns daily and want to move faster without making mistakes
Publishers who manage their own GAM network and want AI-assisted workflows
Media agencies running multiple GAM accounts who want a consistent, auditable process
Developers building AdOps automation on top of Claude, ChatGPT, or Gemini

Supported AI Platforms

| Platform | Setup guide | Mode | | --- | --- | --- | | Claude (Desktop / claude.ai) | docs/install/claude.md | Plugin + MCP remote | | ChatGPT (Pro connector) | docs/install/chatgpt.md | MCP remote (HTTP) | | Gemini | docs/install/gemini.md | Extension + MCP fallback | | OpenAI Codex | docs/install/openai-codex.md | AGENTS.md + MCP wiring |

All platforms connect to the same hosted MCP endpoint at https://orbiads.com/mcp.

8 Built-in Skills

Each skill is a guided, guardrailed workflow. Your AI assistant loads them on demand.

| Skill | What it does | | --- | --- | | bootstrap | Connect your GAM account, verify credentials, select network | | inventory-ad-units | Browse and query ad units, placements, and targeting keys | | availability-forecast | Check inventory availability before trafficking — no writes | | advertiser-order-line-items | Create and manage advertisers, orders, and line items | | placements-targeting | Build and verify targeting configurations | | native-image | Traffic native and image creatives with inline upload | | qa-preview | Preview and coverage check before any live push | | deploy-reporting | Deploy with dry-run protection + post-push delivery reports |

5 Composed Workflows

End-to-end workflows that chain skills automatically.

| Workflow | Description | | --- | --- | | inventory-to-placement | From ad unit discovery to targeting-ready placement | | image-to-native | Upload image → create native creative → traffic line item | | image-to-html5 | Same flow for HTML5 display creatives | | audio-video-trafficking | Audio/video line item setup with format guardrails | | deploy-to-reporting | Push campaign + monitor delivery + generate report |

Safety by Design

Every write action requires explicit confirmation. No campaign goes live by accident.

Dry-run mode on all deployment actions — preview before you push
Forecast gate before inventory commits — availability verified upfront
Audit trail on every action — who did what, when, with what result
Credit guard — read operations are always free, writes deduct credits transparently

Quick Start (3 steps)

1. Create your free account

Go to orbiads.com and sign up. You get 5 free credits — no credit card required.

2. Connect Google Ad Manager

From the OrbiAds dashboard, click Connect GAM and authorize with your Google account. OrbiAds uses OAuth — your GAM credentials never leave Google's infrastructure.

3. Configure your AI assistant

Pick your platform and follow the guide:

Then start with:

"Connect to my GAM account and show me my active networks"

MCP Server Details

| Property | Value | | --- | --- | | Endpoint | https://orbiads.com/mcp | | Transport | streamable-http (default) · sse · stdio (local dev only) | | Auth | OAuth 2.0 — Google account via OrbiAds | | GAM API version | v202602 | | MCP Protocol | 2025-03-26 |

Repository Structure

shared/           ← Canonical skills, agents, prompts, workflows, JSON schemas
docs/             ← Installation guides, safety rules, tool matrix, query library
claude-plugin/    ← Claude plugin packaging (manifest, system prompt, router)
openai-codex/     ← AGENTS.md + router for OpenAI/Codex workspaces
gemini-extension/ ← Gemini extension descriptor + function declarations

Business logic lives in shared/. Platform wrappers are thin adapters that point back to it.

Pricing

| Plan | Price | Credits | | --- | --- | --- | | Trial | Free | 5 credits (no card) | | Starter | €39/month | 50 credits/month | | Early Access | €29/month ← locked for life | 50 credits/month | | Pack S | €29 one-time | +50 credits | | Pack L | €45 one-time | +100 credits |

Reads are always free. Credits are only consumed on write and deploy operations.

Start free →

License

This repository contains the MCP distribution scaffold and platform integration guides for OrbiAds. The OrbiAds MCP server and backend are proprietary — see terms at orbiads.com.

MCP Servers