How MCP Integrations Work: A Complete Guide with Examples
By Samuel Coe, founder of Coe Code • Last Updated: March 2026
The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external data sources and tools using a unified interface. Instead of every AI vendor building custom connectors for every service, services implement MCP once and become accessible to all MCP-compatible AI assistants — including Claude, Cursor, and any compliant client. BrewLogica's MCP server exposes 19 tools across categories for beans, brews, bags, AI features, and utilities. The server is hosted at https://api.brewlogica.app/mcp and uses Bearer token authentication. Setup requires adding a 10-line JSON config to your AI client's configuration file and generating a token at brewlogica.app/tokens. No local server installation is required beyond Node.js 18+ for the npx proxy.
What Is the Model Context Protocol?
MCP defines a standard way for AI assistants to discover and call external tools. At its core, MCP is a client-server protocol where:
- The MCP server (e.g., BrewLogica) exposes a list of tools with name, description, and parameter schema
- The MCP client (your AI assistant) discovers those tools and can call them when the user's request requires it
- The AI assistant decides when and how to call tools based on the conversation — the user doesn't need to specify tool names
This is different from traditional API integrations where each client needs custom code per service. MCP decouples the AI client from the service through a standard protocol, the same way HTTP decouples browsers from web servers.
How BrewLogica's MCP Integration Is Architected
BrewLogica runs a remote MCP server rather than requiring users to run one locally. The connection flow looks like this:
Your Prompt
"Log an espresso: 18g in, 36g out"
AI Assistant
Claude, Cursor, etc.
npx Proxy
mcp-remote (local)
BrewLogica API
api.brewlogica.app/mcp
- 1
Tool Discovery
When your AI client starts, it queries the MCP server for available tools. BrewLogica returns 19 tool definitions including name, description, and JSON Schema for parameters.
- 2
Intent Recognition
When you say "Log an espresso: 18g in, 36g out, 28 seconds", the AI recognizes this requires the
create_brewtool and constructs the appropriate parameters. - 3
Authenticated Tool Call
The AI client calls the tool via the npx proxy, which forwards the request to the BrewLogica server with your Bearer token in the Authorization header.
- 4
Response Processing
BrewLogica creates the brew record and returns a confirmation. The AI presents this to you in a readable format.
All 19 BrewLogica MCP Tools
BrewLogica exposes tools organized into 5 functional categories. The table below shows every tool with its category, a description, and whether it reads or writes data.
| Tool | Category | Description | Type |
|---|---|---|---|
| list_beans | Beans | List all beans in your collection | Read |
| get_bean | Beans | Get details of a specific bean by ID | Read |
| create_bean | Beans | Add a new bean to your collection | Write |
| update_bean | Beans | Update an existing bean's details | Write |
| delete_bean | Beans | Delete a bean from your collection | Delete |
| list_bags | Bags | List all bag purchases for a bean | Read |
| get_active_bag | Bags | Get the currently active bag for a bean | Read |
| create_bag | Bags | Record a new bag purchase with weight and price | Write |
| set_active_bag | Bags | Mark a bag as currently in use | Write |
| list_brews | Brews | List brew history with optional method/bean filter | Read |
| get_brew | Brews | Get details of a specific brew | Read |
| create_brew | Brews | Log a new brew with dose, yield, time, and notes | Write |
| update_brew | Brews | Update an existing brew's rating or notes | Write |
| delete_brew | Brews | Delete a brew from history | Delete |
| parse_brew_text | AI | Parse natural language into structured brew data | AI / Write |
| get_suggestions | AI | Get AI-powered brewing parameter suggestions | AI / Read |
| get_limits | AI | Check AI feature usage and remaining quota | Read |
| get_stats | Utilities | Get brewing analytics for a time period | Read |
| search | Utilities | Search across beans and brews | Read |
Real-World MCP Tool Call Examples
These examples show the natural language prompt, the tool that gets called, and the key parameters the AI constructs automatically.
Prompt
"Log an espresso: 18g in, 36g out, 28 seconds, a touch sour"
Tool Call
create_brew(method: "espresso", dose_grams: 18, yield_grams: 36, time_seconds: 28, notes: "a touch sour")
Prompt
"What were my last 5 V60 brews?"
Tool Call
list_brews(method: "v60", limit: 5)
Prompt
"I just bought a 250g bag of Onyx Tropical Weather, roasted Feb 1st, paid $28"
Tool Calls
1. search(query: "Onyx Tropical Weather") // find existing bean 2. create_bag(bean_id: "...", roast_date: "2026-02-01", bag_weight_grams: 250, price_paid: 28, currency: "USD")
Prompt
"How many brews did I make this month?"
Tool Call
get_stats(period: "month")
How to Connect Your AI Assistant to BrewLogica
- 1 Get a Pro subscription and generate a token at brewlogica.app/tokens
- 2 Add the MCP config to your client's JSON file (see MCP Quickstart for exact config snippets)
- 3 Restart your AI client and say "List my beans in BrewLogica" to verify the connection
Frequently Asked Questions
What is the Model Context Protocol (MCP)?
MCP is an open standard developed by Anthropic that allows AI assistants to connect to external data sources and tools using a standardized interface. Instead of building custom integrations for each AI assistant, services implement MCP once and become accessible to any MCP-compatible client.
What MCP clients support BrewLogica?
BrewLogica's MCP server works with Claude Code, Claude Desktop, Cursor, and any other MCP-compatible client. The server is hosted remotely at https://api.brewlogica.app/mcp, so no local installation is required.
How does BrewLogica authenticate MCP requests?
BrewLogica uses Bearer token authentication. Tokens are generated by the user at brewlogica.app/tokens and passed in the HTTP Authorization header of every MCP request. Tokens start with the bl_ prefix and are scoped exclusively to the account that generated them.
How many MCP tools does BrewLogica expose?
BrewLogica exposes 19 MCP tools across 5 categories: Beans, Bags, Brews, AI Features, and Utilities. See the full tool reference for parameters and usage.
Do I need to install anything locally to use BrewLogica via MCP?
You need Node.js 18 or later to run the npx @anthropic-ai/mcp-remote proxy, which bridges your local MCP client to BrewLogica's remote server. The proxy downloads automatically on first use. The BrewLogica MCP server itself is cloud-hosted.