BridgeCode

The AI coding agent for builders who ship. Multi-provider, multi-agent, terminal-first — with built-in BridgeMind integration.

BridgeCode is a terminal-first AI coding agent built for vibe coding. Write natural language, and BridgeCode handles the execution — reading files, writing code, running commands, searching the web, and orchestrating sub-agents. It ships with multi-provider AI support, custom agent configurations, MCP integration, and runs everywhere: CLI, web UI, and desktop app.

Key Features

• Multi-provider AI — Anthropic (Claude), OpenAI (GPT), Google (Gemini), Groq, Mistral, Amazon Bedrock, OpenRouter, and more. Switch models mid-session.
• Custom agents — Built-in agents for different workflows (vibe, build, plan, architect, ship) plus full support for custom agents via configuration.
• MCP integration — Connect to any MCP server, including BridgeMind's own MCP for project and task management.
• Tool system — File operations, shell execution, web search, code search, LSP integration, sub-agent orchestration, and batch operations.
• Session management — Persistent conversations with compaction, sharing, forking, and export.
• Cross-platform — CLI binary, web interface (SolidJS), and native desktop app (Tauri).

Installation

Quick Install (Recommended)

curl -fsSL https://bridgecode.dev/install | bash

Package Managers

npm i -g bridgecode@latest

bun add -g bridgecode@latest

pnpm add -g bridgecode@latest

docker run -it ghcr.io/anomalyco/bridgecode

Getting Started

Once installed, run BridgeCode in any project directory:

# Start a conversation in the current directory
bridgecode

# Start with a specific prompt
bridgecode "add a user authentication module with JWT"

# Start the web UI
bridgecode web

# Start the TUI (terminal UI)
bridgecode tui

Connect to BridgeMind

BridgeCode ships with built-in BridgeMind MCP integration. Authenticate to access your projects and tasks:

bridgecode auth login

This opens your browser for OAuth authentication with the BridgeMind platform. Once authenticated, BridgeCode can read your projects, pick up tasks, and update progress automatically.

Agents

BridgeCode ships with several built-in agents optimized for different workflows:

Switch agents mid-session with slash commands:

/vibe     Switch to vibe coding mode
/build    Switch to build mode
/plan     Switch to planning mode
/ship     Switch to ship mode

Custom Agents

Define custom agents in your project's bridgecode.json:

{
  "agent": {
    "reviewer": {
      "prompt": "You are a senior code reviewer. Review changes for correctness, performance, and security. Never write code directly — only provide feedback.",
      "temperature": 0.3
    }
  }
}

Tool System

BridgeCode agents have access to a comprehensive set of tools:

File Operations

Search

Execution

External

Permissions

Control which tools agents can use through the permission system:

{
  "permission": {
    "bash": { "allow": true },
    "write": { "*.test.ts": "allow", "*.env": "deny" },
    "read": { "*": "allow" }
  }
}

Configuration

BridgeCode supports multi-layer configuration with the following precedence (highest wins):

1. Inline config — BRIDGECODE_CONFIG_CONTENT environment variable
2. Project config — .bridgecode/config.json or bridgecode.json in the project root
3. Custom path — BRIDGECODE_CONFIG environment variable
4. Global config — ~/.config/bridgecode/config.json
5. Remote config — Well-known URL configs for organizations

Example Configuration

{
  "$schema": "https://bridgecode.dev/config.json",
  "default_agent": "vibe",
  "model": "anthropic/claude-sonnet-4-20250514",
  "provider": {
    "bridgemind": { "options": {} },
    "anthropic": { "options": {} },
    "openai": { "options": {} }
  },
  "mcp": {
    "bridgemind": {
      "type": "remote",
      "url": "https://mcp.bridgecode.dev",
      "enabled": true
    }
  },
  "permission": {
    "read": { "*.env": "deny", "*": "allow" },
    "bash": { "allow": true }
  }
}

Environment Variable Substitution

Reference environment variables in config files:

{
  "provider": {
    "openai": {
      "options": {
        "apiKey": "{env:OPENAI_API_KEY}"
      }
    }
  }
}

MCP Integration

BridgeCode can connect to any MCP server — local or remote:

Local MCP Servers

{
  "mcp": {
    "my-server": {
      "type": "local",
      "command": "npx",
      "args": ["-y", "@my-org/mcp-server"],
      "env": {
        "API_KEY": "{env:MY_API_KEY}"
      }
    }
  }
}

Remote MCP Servers

{
  "mcp": {
    "bridgemind": {
      "type": "remote",
      "url": "https://mcp.bridgecode.dev",
      "enabled": true
    }
  }
}

Remote servers support OAuth authentication for secure access.

Sessions

BridgeCode maintains persistent conversation sessions:

• Auto-compaction — When context fills up, BridgeCode automatically summarizes earlier messages to free space.
• Session sharing — Share sessions with teammates (configurable: manual, auto, or disabled).
• Session forking — Branch a conversation from any message.
• Session export — Export full conversation history.

Web Interface

Launch the web UI for a browser-based experience:

bridgecode web

The web interface provides real-time conversation updates, model selection, agent switching, and full session management — built with SolidJS for performance.

Desktop App

BridgeCode also ships as a native desktop app built with Tauri. Download it from bridgecode.dev for macOS, Windows, and Linux.

Architecture

bridgecode/
├── packages/
│   ├── code/       # Core CLI + server (Hono, AI SDK)
│   ├── app/        # Web UI (SolidJS + Vite)
│   ├── desktop/    # Tauri desktop app
│   ├── console/    # Console web app (SolidStart)
│   ├── ui/         # Shared UI components
│   ├── sdk/        # TypeScript/JavaScript SDK
│   └── plugin/     # Plugin system

The core code package handles all AI interactions, tool execution, session management, and configuration. The app and desktop packages provide frontend interfaces. The sdk package lets you integrate BridgeCode into your own applications.