v1 · 57 tools · OAuth 2.1 · Read-only by default

Adship MCP Server

Connect Claude Desktop, Cursor, or Claude Code to your Adship account and manage Meta, TikTok & Google ads through any MCP-compatible AI agent.

https://mcp.adship.ai/mcp
Install

Pick your client

The same hosted URL works for all of them. OAuth sign-in with your Adship account — no API keys to copy around.

Claude Desktop Recommended

Open Settings → Connectors → Add custom connector and paste:

Name: Adship URL: https://mcp.adship.ai/mcp

Claude Code

One command. After install, run /mcp in Claude Code to trigger OAuth.

claude mcp add --transport http adship https://mcp.adship.ai/mcp

Cursor

Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json (project):

{ "mcpServers": { "adship": { "type": "streamable-http", "url": "https://mcp.adship.ai/mcp" } } }

Claude Desktop (legacy / corp proxy)

For older builds without Custom Connectors. Edit claude_desktop_config.json:

{ "mcpServers": { "adship": { "command": "npx", "args": ["-y", "mcp-remote", "https://mcp.adship.ai/mcp"] } } }
First-run authentication

Sign in once, tokens refresh

The first tool call opens an OAuth window. You sign in to Adship, pick the workspace the agent can access, and approve scopes. Tokens last 24 hours and refresh automatically.

1

Agent makes its first tool call → server returns 401 with OAuth metadata.

2

Agent opens the Adship authorize page in your browser.

3

You sign in (Google / Facebook / email), pick a workspace, click Allow.

4

Agent receives an audience-bound JWT (24h, refreshable, PKCE).

5

Tools work immediately.

Scopes & permissions

Granular, per-platform, revocable

The agent requests scopes; you approve them on the consent screen. Revoke at any time from Adship → Settings → AI Agents.

read:meta

List Meta campaigns, ad sets, ads, creatives & read metrics.

write:meta

Pause / enable / budget / publish ads on Meta.

read:tiktok

List TikTok campaigns, ad groups, ads & read metrics.

write:tiktok

Pause / enable / budget ads on TikTok.

🔒

An agent can never widen its own access. The token is minted with only the scopes you approve, intersected with what the client registered.

Every tool re-checks its scope server-side on each call — a read:meta token cannot pause an ad, touch TikTok, or read your Google Drive. Viewer team members can only obtain read:* — write scopes are stripped at mint, even if requested.

Tools (v1)

57 tools across the workflow

Cross-platform where applicable (Meta + TikTok + Google). The agent calls these directly in your conversation.

🚀 Ad Launching

3 tools
  • bulk_publish
  • bulk_publish_tiktok
  • upload_creative

📋 Campaign Management

13 tools
  • list_campaigns
  • list_ad_sets
  • list_ads
  • duplicate_campaign
  • duplicate_adset
  • create_adset
  • create_tiktok_campaign
  • update_adset_targeting
  • update_budget
  • update_status
  • edit_ads
  • enable_ad
  • pause_ad

📊 Analytics & Reports

11 tools
  • get_account_overview
  • get_ad_metrics
  • get_daily_spend
  • get_top_ads
  • compare_performance_periods
  • get_delivery_status
  • get_budget_pacing
  • get_spend_anomalies
  • get_insights_breakdown
  • get_change_history
  • run_full_account_audit

🎨 Creative Intelligence

5 tools
  • get_creative_concentration
  • group_creatives
  • get_creative_performance
  • generate_ad_copy
  • analyze_creative_strategy

🔭 Competitor Research

3 tools
  • scrape_meta_ads_library
  • analyze_competitor
  • compare_competitors

💬 Comments & Sentiment

4 tools
  • list_comments
  • get_comment_analytics
  • reply_to_comment
  • hide_comment

⚙️ Rules & Recommendations

4 tools
  • list_ad_rules
  • create_ad_rule
  • delete_ad_rule
  • list_recommendations

📁 Drive & Sheets

2 tools
  • browse_google_drive
  • read_google_sheet

🏢 Accounts & Settings

12 tools
  • list_ad_accounts
  • list_workspaces
  • get_current_user
  • get_workspace_settings
  • list_audiences
  • list_pixels
  • list_pages
  • list_creatives
  • list_batches
  • list_conversations
  • get_publish_status
  • delete_entities
Rate limits

Use it respectfully

v1 does not hard-enforce rate limits yet — 429 + Retry-After enforcement ships in v1.1. The published per-token policy:

ClassLimit (v1.1+)
read:*60 calls / min
write:*10 calls / min
bulk_publish3 calls / min (≤ 250 ads each)

New to the Adship connector?

See what you can do with it — launch from Drive, audit accounts, cut waste, spy on competitors.

See the overview →