Agent Bridge MCP Server

A lightweight MCP (Model Context Protocol) server that enables communication between Amazon Quick (local desktop) and Kiro (cloud DevSpaces) through a shared S3 message store.

Architecture

┌──────────┐     ┌─────────────────┐     ┌──────────────┐
│   Kiro   │────►│  Agent Bridge   │◄────│ Amazon Quick │
│(DevSpace)│     │  (S3 bucket)    │     │  (local PC)  │
└──────────┘     └─────────────────┘     └──────────────┘

Both agents run the same server locally (stdio transport). Messages are stored in S3, making them accessible from any environment with AWS credentials.

Related MCP server: AWS S3 MCP Server

Setup

Prerequisites

• Python 3.10+

• AWS credentials with S3 access

• pip install mcp boto3

Environment Variable

export AGENT_BRIDGE_BUCKET=agent-bridge-jnewc

Amazon Quick (Windows)

In Settings → Capabilities → MCP → Add Server:

• Command: py

• Args: C:/Users/jnewc/Desktop/Newcom GitHub repo/agent-bridge-mcp/server.py

• Env: AGENT_BRIDGE_BUCKET=agent-bridge-jnewc

Kiro (DevSpaces)

In .kiro/settings/mcp.json:

{
  "mcpServers": {
    "agent-bridge": {
      "command": "python3",
      "args": ["/workspace/agent-bridge-mcp/server.py"],
      "env": {
        "AGENT_BRIDGE_BUCKET": "agent-bridge-jnewc"
      }
    }
  }
}

Tools (10)

Message Types

• message — General communication

• task — Task handoff with priority and acceptance criteria

• response — Reply to a task or question

• context — Shared context/background info

• status — Status update

• question — Question requiring a response

S3 Structure

bridge/
├── inbox/
│   ├── quick/          # Quick's unread messages
│   └── kiro/           # Kiro's unread messages
└── conversations/
    └── {conv_id}/
        ├── meta.json   # Conversation metadata
        └── messages/   # Chronological messages