Skip to main content
Glama
j-newcom

agent-bridge-mcp

by j-newcom
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Hybrid

Agent Bridge MCP Server

An MCP server + Lambda API that enables real-time communication between Kiro and Amazon Quick across environments (local Windows ↔ cloud DevSpaces). Messages are stored in S3 and accessible via API Gateway — no shared filesystem required.

Architecture

┌──────────────────┐                              ┌──────────────────┐
│   Amazon Quick   │                              │       Kiro       │
│   (Windows PC)   │                              │   (DevSpaces)    │
│                  │                              │                  │
│  MCP Server      │         ┌──────────┐        │  curl / shell    │
│  (stdio, boto3)  │────────►│    S3    │◄───────│                  │
│                  │         │  bucket  │        │                  │
│      — OR —      │         └──────────┘        │      — OR —      │
│                  │              ▲               │                  │
│  Lambda invoke   │              │               │  API Gateway     │
│  (via Matilda)   │────────►┌────┴─────┐◄───────│  (HTTPS POST)    │
│                  │         │  Lambda  │        │                  │
└──────────────────┘         └──────────┘        └──────────────────┘

Related MCP server: mcp-dispatch

Components

Component

Purpose

server.py

Local MCP server (stdio) — Quick connects directly via S3

lambda/lambda_function.py

Lambda function — HTTP API relay to same S3 store

API Gateway

Public HTTPS endpoint for Kiro (DevSpaces can't hit Lambda URLs directly)

S3 Bucket

Shared message store

Endpoints

Who

How they connect

Quick

MCP server (local stdio) with AGENT_BRIDGE_BUCKET env var, OR Lambda invoke via Matilda

Kiro

POST https://<your-api-id>.execute-api.<region>.amazonaws.com/ with x-bridge-secret header

Setup

Prerequisites

  • Python 3.10+

  • pip install mcp boto3

  • AWS credentials with S3 access

Environment Variable

export AGENT_BRIDGE_BUCKET=your-bucket-name

Amazon Quick (local MCP)

Settings → Capabilities → MCP → Add Server:

  • Command: py

  • Args: path to server.py

  • Env: AGENT_BRIDGE_BUCKET=your-bucket-name

Kiro (DevSpaces — via API Gateway)

# Send a message
curl -s -X POST "https://<your-api-id>.execute-api.<region>.amazonaws.com/" \
  -H "Content-Type: application/json" \
  -H "x-bridge-secret: YOUR_SECRET" \
  -d '{"action":"send_message","sender":"kiro","recipient":"quick","content":"Hello from Kiro!"}'

# Check for messages
curl -s -X POST "https://<your-api-id>.execute-api.<region>.amazonaws.com/" \
  -H "Content-Type: application/json" \
  -H "x-bridge-secret: YOUR_SECRET" \
  -d '{"action":"get_messages","agent":"kiro"}'

# Get stats
curl -s -X POST "https://<your-api-id>.execute-api.<region>.amazonaws.com/" \
  -H "Content-Type: application/json" \
  -H "x-bridge-secret: YOUR_SECRET" \
  -d '{"action":"get_stats"}'

API Actions

All requests are POST with JSON body containing "action" plus parameters:

Action

Parameters

Description

send_message

sender, recipient, content, conversation_id?, title?, message_type?

Send a message

get_messages

agent, status? (unread/read/all), limit?

Check inbox

mark_read

agent, message_ids?, conversation_id?

Mark messages read

list_conversations

status? (active/closed/all)

List threads

get_conversation

conversation_id

Full thread history

get_stats

(none)

Bridge statistics

Message Types

message, task, response, context, status, question

S3 Structure

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

Security

  • S3 bucket: private, all public access blocked

  • API Gateway: secured by x-bridge-secret header checked in Lambda code

  • Lambda function URL: AuthType: NONE (backup path)

AWS Resources (you'll create these)

  • S3 Bucket: any name (set as AGENT_BRIDGE_BUCKET)

  • Lambda: handles API actions against the S3 bucket

  • API Gateway: HTTP API fronting the Lambda

  • IAM Role: Lambda execution role with S3 read/write on the bucket

License

MIT

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityInactive
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers