Model Context Protocol

Use AdRevila from your favourite AI agent.

We speak the Model Context Protocol — the open standard agent clients (Claude Desktop, Cursor, ChatGPT, Copilot Chat, Zed, every modern MCP-aware tool) use to talk to external services. Three tools, one bearer token. No SDK to install.

Set up the integration

Configs for Claude Desktop · Cursor · ChatGPT Custom GPTs

What is MCP?

The Model Context Protocol is a JSON-RPC standard that lets an AI agent call external services as tools. Instead of building a custom integration for every model + every service, agent clients implement MCP once and gain access to anything that speaks it.

Anthropic released it in late 2024. By 2026 it's the default — Claude Desktop, Cursor, ChatGPT, VS Code Copilot Chat, Zed, Cline, Continue, and most agent frameworks ship MCP support out of the box. GitHub, Linear, Figma, Cloudflare, Stripe, Notion, Slack, and Sentry all run MCP servers. So do we.

modelcontextprotocol.io ↗

What you can do with it.

Three tools. Same surface a developer hits over plain HTTP, packaged so your agent can call it in plain English.

analyze_ad

"Use adrevila to analyze this URL and summarize the hook."

The agent submits the URL, waits for the worker (up to 90s), and reads the report back.

list_analyses

"Show me my last 10 AdRevila teardowns, sorted by score."

The agent fetches your completed analyses, filters and ranks in-memory.

get_analysis

"Pull the markdown for that TJ Maxx report and email it to my creative team."

The agent fetches the report as Markdown, hands it to whichever tool composes the email.

Why this matters.

Most apps that expose APIs assume you'll write code against them. That assumption is breaking down. A real slice of API traffic in 2026 comes from AI agents acting on natural-language prompts. If your tool can't be reached that way, you're invisible to the workflows your customers are already building.

We speak MCP for the same reason we ship a REST API, an OpenAPI spec, and a /llms.txt: different audiences reach AdRevila through different protocols, and we'd rather meet them where they already work than make them adopt ours.

How it works.

01
Create a token
Sign in, go to Settings → Developers, and create a personal access token. Same token works for REST and MCP — one secret, all surfaces.
02
Point your client at the endpoint
Add https://adrevila.com/api/v1/mcp to your agent client's MCP config. Two-line JSON snippet, one Authorization header.
03
Prompt naturally
Your agent now sees three AdRevila tools. Ask it questions like the examples above; it orchestrates the calls. No manual API knowledge required.

Full setup guide

FAQ

Do I need a separate plan for MCP access?

No. Same credits, same tokens, same rate limits. The MCP endpoint has its own per-token rate-limit bucket (60/min) so agent traffic doesn't deplete your REST quotas.

Which clients work?

Anything that supports MCP over HTTP. That covers Claude Desktop, Cursor, ChatGPT (since early 2026), VS Code Copilot Chat, Zed, Cline, Continue, and most agent frameworks (Mastra, OpenAgents, LangChain MCP adapter, etc.).

Is there an npm package or stdio version?

Not yet. The HTTP transport covers every canonical client today. We'll publish a stdio version if real users ask for it.

What about webhooks?

Not yet. MCP is request-driven; the agent calls when it wants data. Webhooks (push-style notifications) are on the v4 list.

Can I see the spec?

Yes — /api/v1/openapi.yaml documents the REST surface; /developers/agents walks through the MCP tools and configs.

Plug AdRevila into your agent stack.

Same tokens as the REST API. Three tools. No SDK.

Set it up →