Smart AI Bridge

Smart AI Bridge MCP Server enables seamless orchestration of multiple AI backends for Claude Code. It sits between Claude Code and your AI providers, offering a configurable council system, smart routing, and a suite of tools to optimize code-related tasks and workflows.

How to use

You use an MCP client to connect to the Smart AI Bridge MCP server. Start by ensuring at least one backend is configured, then add the bridge as an MCP server entry in Claude Code. Once connected, you gain access to all 19 tools and can tailor routing, council strategies, and topic mappings to your needs. Use the council features to query multiple backends on the same prompt and have Claude synthesize the responses.

How to install

Prerequisites: install Node.js version 18 or newer.

# 1. Install dependencies for the Smart AI Bridge MCP Server
cd /path/to/smart-ai-bridge
npm install

# 2. Ensure you have a configured backend before starting
# Backend configuration lives in src/config/backends.json

# 3. Add the MCP server entry to Claude Code (see installation steps below)

Next steps to run and verify

Start the MCP server from its local directory using the exact runtime command shown in the integration example.

node src/server.js

Configure and connect to Claude Code

Configure Claude Code to recognize the bridge by adding an MCP server entry with the provided runtime details. Use the following snippet exactly as shown to connect from Claude Code.

{
  "mcpServers": {
    "smart_ai_bridge": {
      "command": "node",
      "args": ["src/server.js"],
      "cwd": "/path/to/smart-ai-bridge",
      "env": {
        "NVIDIA_API_KEY": "your-key",
        "OPENAI_API_KEY": "your-key",
        "GEMINI_API_KEY": "your-key",
        "GROQ_API_KEY": "your-key"
      }
    }
  }
}

Available tools

analyze_file

Reads and analyzes a file, returning structured findings to help understand its content and quality.

modify_file

Applies natural-language edits to a file and returns the resulting diff.

batch_analyze

Analyzes multiple files matched by glob patterns, returning aggregated findings.

batch_modify

Applies the same set of instructions to multiple files and returns per-file results.

generate_file

Generates code from a natural-language specification.

explore

Answers questions about the codebase using intelligent search over the repository contents.

ask

Smart routing with auto or forced backend selection for a given prompt.

council

Queries multiple backends and returns all responses for synthesis.

dual_iterate

Generates, reviews, and fixes code across two backends in a loop.

parallel_agents

Runs a Test-Driven Development workflow with decomposition and quality gates across agents.

spawn_subagent

Creates specialized AI agents for focused tasks like TDD or architecture work.

review

Performs security, performance, and quality assessments on code changes.

refactor

Refactors code across multiple files while updating references.

validate_changes

Pre-flight validation of proposed changes before applying them.

check_backend_health

Measures health and readiness of a specific backend.

backup_restore

Manages timestamped backups and restores state as needed.

write_files_atomic

Writes multiple files atomically with built-in backups.

manage_conversation

Keeps a coherent multi-turn conversation thread across prompts.

get_analytics

Provides usage analytics and optimization recommendations.