# AdsGPT

Hooks your AI client up to ad accounts through AdsGPT: campaign plans, ad copy, reports and budget edits. You approve every change yourself.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: adsgpt-dev
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/adsgpt-dev/agent-plugins/tree/HEAD/plugins/adsgpt
- Updated: 2026-10-21
- HTML: https://affpapa.org/en/nevibecode/mcp/adsgpt

## What it does

AdsGPT is a remote MCP server at `https://mcp.adsgpt.dev`. The client signs in with OAuth 2.1, asks for `ads:read`, and requests `ads:write` the first time a tool changes something. Nothing runs locally — the package ships no scripts, hooks or binaries, just Markdown skills, per-client manifests and the server address.

Tools split into reads and writes. `list_accounts` shows what is connected, `google_ads_report` and `google_ads_search` return performance and GAQL queries, while `google_ads_fields` and `google_ads_docs` cover field lookup and the API reference. Writes go through `save_blueprint`, `save_creatives`, `save_image_creatives`, `save_video_creatives`, `launch_blueprint`, `google_ads_update_budget`, `google_ads_update_status` and `google_ads_add_negative_keywords`. A new campaign starts paused, and the client asks for approval before each tool that modifies anything.

You need an AdsGPT account with an ad account attached. The `google_ads_*` tools only show up once your organization has connected Google Ads. What reaches the server is the tool call input: account id, GAQL query, campaign plan, ad texts, image and video links, landing page URL. Your conversation stays out of it. The skills (`adsgpt-agent`, `adsgpt-create-campaign`, `adsgpt-google-ads`, `adsgpt-write-ads`, `adsgpt-optimize`) are optional for the server to work, but they push the model to ask the right questions instead of guessing.

## Requirements
- Remote server: `https://mcp.adsgpt.dev`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http adsgpt https://mcp.adsgpt.dev
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add adsgpt --url https://mcp.adsgpt.dev
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "adsgpt": {
            "url": "https://mcp.adsgpt.dev"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "adsgpt": {
            "type": "http",
            "url": "https://mcp.adsgpt.dev"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http adsgpt https://mcp.adsgpt.dev
```

## When to use it
- build a campaign plan and check it against platform rules
- write ad copy that fits each format's limits
- pull spend and results for campaigns, ads or keywords
- lower a campaign budget or pause it
- add negative keywords to a campaign

## Example prompts
- List the connected ad accounts and the platforms I can add
- Draft a campaign plan for signups from `landing.html` and show it for approval
- Report last week's spend and suggest where to cut the budget

## Similar
- [google-ads-report](https://affpapa.org/en/nevibecode/skills/google-ads-report.md) — skill · 476 installs
- [Adwarp](https://affpapa.org/en/nevibecode/mcp/adwarp.md) — MCP server · ★ 23
- [Google Ads](https://affpapa.org/en/nevibecode/mcp/google-ads.md) — MCP server
- [LinkedIn Ads](https://affpapa.org/en/nevibecode/mcp/linkedin-ads.md) — MCP server
- [adspy-analytics-intelligence](https://affpapa.org/en/nevibecode/skills/adspy-analytics-intelligence.md) — skill · 756 installs
- [amazon_ads_mcp](https://affpapa.org/en/nevibecode/mcp/amazon-ads-mcp.md) — MCP server · ★ 74

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md