Skip to content

README.md

Latest commit

 

History

History
111 lines (82 loc) · 4.06 KB

README.md

File metadata and controls

111 lines (82 loc) · 4.06 KB

Publish TiDB MCP Server to AWS Marketplace

This folder contains the baseline assets to package this repository's MCP server as an AWS Marketplace AI tool.

Requirements

  • Python >= 3.10
  • MCP Python SDK >= 1.26.0 (installed via pip install ".[mcp]")

What this package does

  • Runs tidb-mcp-server in streamable-http mode
  • Binds to 0.0.0.0:8000
  • Exposes MCP endpoint at POST /mcp
  • Uses stateless HTTP with MCP-compatible responses (application/json or text/event-stream)

1) Build the container image

Ensure you have the MCP dependencies available (defined in pyproject.toml under [project.optional-dependencies]):

docker build -f deploy/aws-marketplace/Dockerfile -t tidb-mcp-server:marketplace .

2) Run a local smoke test

docker run --rm -p 8000:8000 \
  -e TIDB_HOST='<your-tidb-host>' \
  -e TIDB_PORT='4000' \
  -e TIDB_USERNAME='root' \
  -e TIDB_PASSWORD='' \
  -e TIDB_DATABASE='test' \
  tidb-mcp-server:marketplace

The container runs with --transport streamable-http --host 0.0.0.0 --port 8000 by default. /mcp path and stateless mode are fixed for predictable Marketplace behavior.

In another terminal, call the MCP endpoint:

# Initialize the MCP session
curl -i -X POST http://localhost:8000/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{
    "jsonrpc": "2.0",
    "id": "1",
    "method": "initialize",
    "params": {
      "protocolVersion": "2025-03-26",
      "capabilities": {},
      "clientInfo": {"name": "smoke-test", "version": "0.1.0"}
    }
  }'

The server will return MCP-compliant JSON responses.

3) Push image to Amazon ECR

AWS_ACCOUNT_ID='<your-account-id>'
AWS_REGION='us-east-1'
ECR_REPO='tidb-mcp-server'

aws ecr create-repository --repository-name "${ECR_REPO}" --region "${AWS_REGION}" || true
aws ecr get-login-password --region "${AWS_REGION}" \
  | docker login --username AWS --password-stdin "${AWS_ACCOUNT_ID}.dkr.ecr.${AWS_REGION}.amazonaws.com"

docker tag tidb-mcp-server:marketplace "${AWS_ACCOUNT_ID}.dkr.ecr.${AWS_REGION}.amazonaws.com/${ECR_REPO}:v1"
docker push "${AWS_ACCOUNT_ID}.dkr.ecr.${AWS_REGION}.amazonaws.com/${ECR_REPO}:v1"

4) Create the AWS Marketplace listing

In AWS Marketplace Management Portal, create an AI Agents and Tools product with Container-based AI agent delivery.

Use these runtime values:

  • Protocol: MCP streamable-http
  • Mode: stateless
  • Endpoint path: /mcp
  • HTTP method: POST
  • Container listen address: 0.0.0.0:8000

Then provide:

  • Container image URI from ECR
  • Deployment template/runtime details (typically ECS/EKS)
  • Usage instructions and required environment variables (TIDB_HOST, TIDB_PORT, TIDB_USERNAME, TIDB_PASSWORD, TIDB_DATABASE)

5) Pre-submission checklist

  • The image starts with no manual bootstrap steps.
  • POST /mcp is reachable and returns valid MCP responses.
  • Service runs in stateless mode.
  • Documentation explains required TiDB credentials and network access.
  • Metering/pricing dimensions are configured in the listing.

References