Model Context Protocol

MCP Server

Connect GetHookd to Your AI Workflow

The GetHookd MCP server lets Claude Desktop, Claude Code, Cursor, and any Model Context Protocol-compatible client query your ad intelligence in plain language - no custom REST integration, no glue code, no context switching. Point your AI tool at our endpoint, authenticate with your API token, and start asking questions like "show me the top-performing ads for this brand" directly inside your editor.

Start 7-Day Free Trial
Read the docs

Plug Gethookd Into Any AI Workflow

Connect Claude Desktop, Claude Code, Cursor, or any Model Context Protocol-compatible AI client to your Gethookd workspace and start exploring millions of ads in plain language.

1

Generate an API token

In your Gethookd account, go to Settings → API Tokens and create a new token with the scopes you want Claude to use.

Open API Tokens
2

Add Gethookd to your config

Open claude_desktop_config.json and paste the Gethookd MCP server entry. Use this URL:

https://app.gethookd.com/api/mcp/v1
3

Restart and ask Claude

Fully quit and relaunch Claude Desktop, then try a prompt like "List the brands I'm spying on." Claude calls the right MCP tool and returns the answer.

Pair with Claude

Use Gethookd + Perplexity together via Claude

Perplexity ships its own MCP server (web search + reasoning), but doesn't consume external MCP servers as a client. To use Gethookd's ad-library tools and Perplexity's web-search tools in one conversation, point Claude Desktop at both: add Gethookd as one MCP server and Perplexity as another. Claude routes each prompt to whichever tool fits.

Set up Claude Desktop first
1

Generate an API token

In Gethookd, go to Settings → API Tokens and create a token with the scopes your OpenClaw agents will use. OpenClaw runs 24/7, so treat the token like a service credential.

Open API Tokens
2

Add Gethookd to OpenClaw's config

Edit ~/.openclaw/config.json and add a mcp.servers.gethookd entry. OpenClaw uses the streamable-http transport for hosted MCP servers:

https://app.gethookd.com/api/mcp/v1
3

Or add it via CLI

Skip the JSON edit and let the CLI write it for you. After, restart the OpenClaw service so the agent picks up the new tools.

Read the full guide
1

Generate an API token

In Gethookd, go to Settings → API Tokens and create a token. Hermes Agent reads tokens from its YAML config, not from the environment, so paste-only access is fine.

Open API Tokens
2

Add Gethookd to ~/.hermes/config.yaml

Add a mcp_servers.gethookd entry with our hosted URL and the bearer token under headers. Hermes handles HTTP MCP servers natively:

https://app.gethookd.com/api/mcp/v1
3

Reload without restarting

Inside any Hermes chat, run /reload-mcp. The agent picks up the Gethookd tools immediately, no process restart required.

1

Generate an API token

NemoClaw wraps OpenClaw inside an NVIDIA OpenShell sandbox, so token handling lives at the OpenClaw layer. Create a Gethookd token with the scopes your sandboxed agent needs.

Open API Tokens
2

Configure inside the sandbox

NemoClaw mounts a per-agent OpenClaw config inside the OpenShell container. Add Gethookd to that file (mcp.servers.gethookd) - same JSON shape as standalone OpenClaw, same streamable-http transport.

https://app.gethookd.com/api/mcp/v1
3

Allow egress to app.gethookd.com

OpenShell blocks outbound network by default. Approve the app.gethookd.com host in the NemoClaw policy console (or pre-allow it in your sandbox manifest) so the MCP request can leave the sandbox. The sandbox redacts your token from logs automatically.

Tip Using Claude Code? Run /mcp after restart to list every connected server and its tools.
Once you're connected

What you can ask Claude

Real prompts that work the moment Gethookd is wired into Claude. No custom scripts, no API plumbing - just chat.

Show me the top 5 fitness brands running Facebook ads right now.

Claude calls search_brands

Returns a ranked brand list - names, ad counts, performance signals - Claude can drill into any of them with one follow-up.

What ads is Gymshark running this month?

Claude calls get_brand

Pulls Gymshark's active ads with creatives, headlines, hook scores, and platform breakdown - straight into the chat.

Find 10 Instagram ads with strong hooks in the beauty space.

Claude calls search_ads

Filters by platform, vertical, and hook quality; returns ad cards with thumbnails and copy ready for swiping.

Which brands in my swipe file added new ads this week?

Claude calls list_brand_spies

Walks your tracked brands, flags the ones with new creative since last check, and lists what changed.

Get the full creative for ad 5821794 and explain why it works.

Claude calls get_ad

Fetches the full ad payload - copy, visuals, CTA, performance - so Claude can break down the angle and structure.

Pick 3 winning dropshipping ads I haven't seen yet.

Claude calls search_ads

Combines bucket filters with your view history to surface fresh winners - no scrolling, no duplicates.

Seven tools available out of the box. See the full tool reference.

Frequently Asked Questions About the GetHookd MCP Server

What is MCP and how does it work with GetHookd?
The Model Context Protocol (MCP) is an open standard that lets AI tools like Claude Desktop, Claude Code, and Cursor call external data sources through structured tools. The GetHookd MCP server exposes your ad intelligence as seven callable tools - search_ads, get_ad, get_brand, search_brands, list_brand_spies, get_brand_spy, and get_top_ads - so your AI agent can query GetHookd in plain language without you writing a single line of API integration code.
Which AI tools are compatible with the GetHookd MCP server?
Any client that implements the Model Context Protocol can connect. Confirmed clients include Claude Desktop (macOS, Windows, Linux), Claude Code, and Cursor. The server also speaks plain HTTP, so any tool that can issue authenticated POST requests to the /api/mcp/v1/call endpoint can invoke tools directly - including custom scripts and automation workflows.
Do I need a paid plan to use the MCP server?
Yes. Public API access - which the MCP server uses - is gated by the public_api feature flag, available on paid plans. See our Pricing page for current plan details. Once you're on a qualifying plan, you can generate an API token in Settings → API Tokens and use it to authenticate the MCP server.
How do I update the MCP server when GetHookd ships new tools?
Nothing to update on your side for non-breaking additions. The server endpoint (https://app.gethookd.com/api/mcp/v1) is versioned - new tools and properties are added within /v1 without breaking existing clients. The tool manifest at GET /api/mcp/v1/tools always reflects the current list. Breaking protocol changes ship under /v2 with advance notice.
Is my API token and data secure?
Yes. All traffic is encrypted over HTTPS. Your API token is scoped to the permissions you select - for example, explore:read grants ad search access only, while brand-spy:read grants brand tracking access. Tokens can be revoked at any time in Settings → API Tokens. The MCP server never stores your prompts or conversation history - it only processes the structured tool call your AI agent sends.
Does each MCP tool call cost credits?
Search tools (search_ads, list_brand_spies, get_brand_spy) use the same credit system as the REST API - an MCP-driven search and a direct curl request cost the same. Lookup tools (get_ad, get_brand, search_brands, get_top_ads) are free. Every charged response includes used_credits and remaining_credits in the envelope so you always know where you stand. Check Settings → API Tokens for your plan's current credit rate.
What happens if my AI agent makes too many requests?
The MCP server enforces the same per-token rate limit as the REST API. If you exceed it, the server returns 429 Too Many Requests. The limit resets on a sliding window - back off briefly and retry. Compliant MCP clients like Claude Desktop and Claude Code handle reconnects automatically.
Can I use the same API token for both MCP and the REST API?
Yes - one token authenticates both surfaces. Generate a key in Settings → API Tokens, pick the scopes you need, and use the same Bearer token in your MCP config and your REST API Authorization header. This is by design: the MCP server is a structured tool layer over the same /api/v1/* endpoints the REST API exposes.
Ready to scale smarter? 🚀
Streamline your creative process, stabilize ROAS, and unlock profitable growth-without the guesswork.
Start your free trial
All-in-one platform that centralizes research, planning, production, and analysis of your ad creatives.
Sitemap
Affiliate
new
Blog
Connect
[email protected]
LinkedIn
YouTube
Instagram
More from GetHookd
The AI Ad Generator Built to Create Winning Ads 3X Faster
AI Creative Analytics: Unlock Creative Performance With Advanced Analytics
Chrome Extension
Download
Chrome logo
Chrome Extension
Refund Policy
Privacy Policy