ardhiee/lamtat-mcp-server

lamtat-mcp-server

A tiny FastMCP server exposing two example tools — process_data (resource fetch + summary) and multiply (numerical product) — configured for stateless HTTP transport.

Code

from fastmcp import FastMCP, Context

mcp = FastMCP(
    name="lamtat-mcp-server",
    host="0.0.0.0",
    port=6565,
    stateless_http=True,
    json_response=True,
)

@mcp.tool
async def process_data(uri: str, ctx: Context):
    await ctx.info(f"Processing {uri}...")
    data = await ctx.read_resource(uri)
    summary = await ctx.sample(f"Summarize: {data.content[:500]}")
    return summary.text

@mcp.tool
def multiply(a: float, b: float) -> float:
    return a * b

mcp.run("streamable-http")

The repository keeps a thin wrapper (server.py) so you can ship the server as a module or container and configure host/port via environment variables.

store_docs tool

Use store_docs to upload one or more documents, store them under raw/<team>/ in S3, then chunk, embed with Bedrock Cohere, and index those chunks into OpenSearch:

@mcp.tool
async def store_docs(team: str, files: list[dict[str, Any]], ctx: Context):
    ...

Provide a files list; for single-file uploads supply a list with one {"filename", "content_base64"} entry (optionally content_type, uploaded_by, allowed_teams, tags, repo, path). Each chunk is embedded via the Bedrock Cohere model and written to the configured OpenSearch index.

Required environment variables:

• S3_BUCKET_NAME
• AWS_REGION (or AWS_DEFAULT_REGION)
• BEDROCK_REGION (defaults to AWS_REGION)
• BEDROCK_MODEL_ID
• OPENSEARCH_ENDPOINT
• OPENSEARCH_INDEX
• Optional: CHUNK_SIZE, CHUNK_OVERLAP

Run locally

python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -e .
lamtat-mcp-server

Environment variables:

• APP_NAME – display name (default lamtat-mcp-server)
• HOST – bind address (default 0.0.0.0)
• PORT – listen port (default 6565)

Container

docker build -t lamtat-mcp-server:latest .
docker run --rm -p 6565:6565 lamtat-mcp-server:latest

Deploy with AWS Copilot

1. Ensure the Copilot CLI is installed and AWS credentials are configured.
2. Initialise (skip if already done):

   copilot init --app lamtat-mcp --name lamtat-mcp-server --type "Load Balanced Web Service" --dockerfile Dockerfile
   

3. Deploy, e.g. to test:

   copilot env init --name test --profile default --region us-east-1
   copilot deploy --name lamtat-mcp-server --env test
   

The manifest at copilot/lamtat-mcp-server/manifest.yml exposes port 6565 with the default FastMCP routing.

Copilot scripts

• scripts/deploy.sh – bootstrap app/env (if needed) and deploy service
• scripts/destroy.sh – remove service, env, and app

For ECS on x86 (Fargate), ensure images are built for linux/amd64. Copilot handles this via the manifest platform setting; for manual builds use docker build --platform linux/amd64 ....