Get started

AI agents, MCP, and Brew — a beginner's guide.

If you've heard about AI agents and MCP but aren't sure how they fit together — or where Brew comes in — this guide is for you.

What are AI agents?

An AI agent is an AI model that can take action — not just answer questions, but actually do things. Instead of giving you a recipe, an agent can order the groceries. Instead of writing a draft email, it can send it.

Agents work by using tools. Each tool is a capability the agent can call — search the web, send a message, generate an image, query a database. The agent decides which tool to use and when, based on what you ask it to do.

Example

“Find me a cheap flight to Tokyo next week and add it to my calendar.”

The agent searches for flights (tool 1), picks the best option, then creates a calendar event (tool 2).

What is MCP?

MCP stands for Model Context Protocol — an open standard that lets AI agents discover and use tools from external services. Think of it as a USB-C for AI agents: one universal connector that works with any agent and any service.

Before MCP, every agent framework had its own way of connecting to tools. If you built a service, you had to write separate integrations for Claude, ChatGPT, Copilot, Cursor, and every other agent. With MCP, you write one integration and every MCP-compatible agent can use it.

Without MCP

Build a custom plugin for every agent platform. Repeat for each service.

With MCP

Build one MCP server. Every MCP-compatible agent can use it instantly.

How agents and MCP work together

Here's the flow when you ask an agent to do something that involves an external service:

You: "Generate a product image for my new sneaker design."

Agent (Claude, GPT, Copilot, etc.)
  │
  ├─ 1. Understands your goal
  ├─ 2. Checks available MCP tools
  ├─ 3. Picks "generate-image" from an image-generation MCP server
  ├─ 4. Calls the MCP tool with your prompt
  │
  MCP Server (e.g. an image generation API)
  │
  ├─ 5. Receives the request
  ├─ 6. Generates the image
  └─ 7. Returns the result to the agent

Agent: "Here's your sneaker design image!"

The agent handles the reasoning. MCP handles the connection. You just get the result.

The problem: too many services, no way to compare

As more services launch MCP servers, a new problem emerges: how do you find the right one?

There are dozens of MCP servers for image generation, email, search, data analysis — and more every week.
Which one is the fastest? The cheapest? The most reliable? The safest?
How do you know a service is trustworthy before you let your agent use it?
How do you set budgets so your agent doesn't accidentally spend too much?

How Brew helps

Brew is the discovery, trust, and procurement layer for MCP services. It helps both humans and autonomous agents find, compare, and safely use MCP servers.

Discover

Search Brew Registry to find MCP servers by capability. Need image generation? Brew shows you every option ranked by relevance, cost, and quality.

Compare

Compare providers side by side — pricing, latency, reliability scores, trust ratings. Brew does the research so your agent doesn't have to.

Trust

Every provider on Brew has a trust score, risk classification, and permission analysis. Know what data a service accesses and how risky it is before you connect.

Procure

Brew recommends the best provider for your specific task. Your agent can ask Brew directly: 'What's the best MCP server for sending transactional emails?' and get an instant, data-driven answer.

Govern

Set budgets, require verified providers, block high-risk actions. Brew gives you control over what your agents can do and how much they can spend.

Try it yourself

Here are a few things you can do right now:

Browse the Registry

See what MCP servers are already listed on Brew.

Browse capabilities →

Publish your own MCP server

Have an MCP service? Submit it to Brew Registry.

Publish a provider →

Read the docs

Deep dive into Brew's architecture, API, and MCP tools.

View documentation →

Configure your agent

Connect your AI agent to Brew and let it discover providers automatically.

See config examples →

Quick glossary

AI Agent: An AI model that can use tools to take actions — not just generate text.
MCP (Model Context Protocol): An open standard that lets AI agents discover and call tools from external services. One integration works with every MCP-compatible agent.
MCP Server: A service that exposes capabilities (tools) via the MCP standard. Agents connect to MCP servers to do things like send emails, generate images, or query databases.
Brew Registry: A directory of MCP servers with search, comparison, trust scoring, and procurement recommendations.
Trust Score: Brew's rating of how safe and reliable an MCP provider is, based on permissions, data sensitivity, verification status, and risk classification.