# Agent-Ready Bridge

> Agent-Ready Bridge is an open-source Model Context Protocol (MCP) server that gives AI assistants read-only access to a small business's existing data (appointments, customers, orders) stored as local JSON. Tagline: "Less busywork. More business."

Agent-Ready Bridge runs locally over stdio. It makes no network calls and never writes to the data file. The operator chooses the data file through the BUSINESS_DATA_PATH environment variable; the assistant cannot choose which file is read.

## Capabilities

- Tool `fetch_business_data`: read-only. With no arguments it lists collections and record counts. With `collection` it returns records, optionally narrowed by exact-match, case-insensitive `filters`, and capped by `limit` (default 50, maximum 200).
- Transport: stdio. Runtime: Node.js 18 or newer.
- Data format: a JSON object whose keys are collections, each an array of records. The bundled sample, `database.json`, holds `appointments` with `clientName`, `appointmentTime`, `serviceType` and `status`.
- Compatible with MCP-capable clients such as Claude Desktop and Claude Code. Other MCP clients can use the same command and arguments.

## Implementation

- Step 1: Connect. Securely point the AI bridge at your existing local office spreadsheets.
- Step 2: Connect the Pieces. We configure a quiet translation script right inside your office computer that turns your messy everyday files into plain English your AI can read.
- Step 3: Deploy. Your team can instantly talk to your office files like a private chat message inside your AI window.
- Full walkthrough: [How it works](how-it-works.html).
- Implementation requests are submitted through the intake form on the [storefront](index.html#intake).

## Commercial Pricing Matrix

All services are custom and hand-delivered per contract. Rates are in USD.

### Tier 1: The Administrative Leak Audit, $199 (one-time)

- Locate the wasted employee hours hiding in your admin routines
- Pinpoint repeated lookups, re-entry and manual follow-ups that cost time every week
- Clear list of where the hours go, so you know what to fix first

### Tier 2: The Core Office AI Integration, $950 (one-time)

- Local file query setup: ask questions about your own scheduling and records files
- Text-based data queries: ask in plain English, get plain-text answers
- Zero cloud overhead: the bridge runs on your own computer, with no cloud server to host or pay for
- Office privacy: your files stay on your own computer

### Tier 3: The AI Search Engine Optimization Engine, $99/mo (monthly utility)

- Ongoing low-friction script maintenance
- Background semantic data updates (llms.txt schema mapping)
- Keeps your site readable to AI search tools, such as ChatGPT Search, so they can find and recommend you

## Docs

- [Repository and README](https://github.com/doncelettabellamy-create/agent-ready-bridge): setup, tool parameters, client configuration and security notes.
- [Server source](https://github.com/doncelettabellamy-create/agent-ready-bridge/blob/main/index.js): the full implementation in a single file.
- [Sample data](https://github.com/doncelettabellamy-create/agent-ready-bridge/blob/main/database.json): example appointments dataset.
- [OpenAPI specification](openapi.yaml): route definition for the read-only `GET /api/fetch-records` gateway, with `clientName`, `totalStaffHoursLost` and `dataType` query parameters. Preview only: the hosted gateway is simulated and not live yet.

## Integration

- Install: `npm install`, then `npm start` from the repository folder.
- Claude Code: `claude mcp add agent-ready-bridge -e BUSINESS_DATA_PATH=/path/to/data.json -- node /path/to/agent-ready-bridge/index.js`
- Claude Desktop: add an `mcpServers` entry with `command: "node"` and the path to `index.js`.

## Optional

- [Storefront](index.html): human-readable overview of the product.