Deployment

Deployment Options

audio-transcriber supports local stdio, a loopback-only development listener, a least-privilege stdio container, and a remote authenticated HTTPS boundary. Provider endpoint, credential, selector, identity, and trust material are supplied at runtime through AgentConfig; none is stored in this repository.

Installed stdio process

{
  "mcpServers": {
    "audio-transcriber": {
      "command": "audio-transcriber-mcp",
      "args": [],
      "env": {"MCP_TOOL_MODE": "intent"}
    }
  }
}

Loopback development listener

audio-transcriber-mcp --transport streamable-http --host 127.0.0.1 --port 8000

Do not expose this listener beyond loopback. Network deployments require direct TLS or an explicitly trusted TLS-terminating ingress, configured authentication, exact MCP_ALLOWED_HOSTS, and an exact trusted-proxy CIDR policy.

Least-privilege local container

docker run -i --rm \
  --read-only \
  --cap-drop=ALL \
  --security-opt=no-new-privileges \
  --pids-limit=256 \
  --tmpfs /tmp:rw,noexec,nosuid,nodev,size=64m \
  -e TRANSPORT=stdio \
  registry.example.invalid/audio-transcriber@sha256:<digest> audio-transcriber-mcp

The operator projects the selected AgentConfig profile into the process at runtime; the image remains immutable and contains no environment connection profile.

Remote authenticated HTTPS endpoint

{
  "mcpServers": {
    "audio-transcriber": {"url": "https://service.example.invalid/mcp"}
  }
}

Store the real remote URL, outbound identity reference, and TLS-profile reference in AgentConfig, not in MCP client JSON or documentation.

This page covers running audio-transcriber as a long-lived server: the transports, proxy, and giving it a DNS name with Technitium.

audio-transcriber ships an MCP server (console script audio-transcriber-mcp) server is the typed, deterministic tool surface; the agent drives those tools over the Agent Control Protocol.

Run the MCP server

The transport is selected with --transport (or the TRANSPORT env var):

stdio (default)streamable-httpsse

audio-transcriber-mcp

For IDE / desktop MCP clients that launch the server as a subprocess.

audio-transcriber-mcp --transport streamable-http --host 0.0.0.0 --port 8000

A network server with a /health endpoint and /mcp route.

audio-transcriber-mcp --transport sse --host 0.0.0.0 --port 8000

Health check (HTTP transports):

curl -s http://localhost:8000/health        # {"status":"OK"}

Configuration (environment)

audio-transcriber is configured from the environment. The commonly used set:

Var	Default	Meaning
HOST	0.0.0.0	Bind address for HTTP transports
PORT	8000	Bind port for HTTP transports
TRANSPORT	stdio	stdio, streamable-http, or sse
WHISPER_MODEL	base	Whisper model: tiny, base, small, medium, large
TRANSCRIBE_DIRECTORY	data dir	Default directory for recordings and exports
AUDIO_PROCESSINGTOOL	True	Register the audio-processing tool set
MISC_TOOL	True	Register the miscellaneous (health) tool set
ENABLE_OTEL	True	Export OpenTelemetry traces
EUNOMIA_TYPE	none	Authorization mode: none, embedded, remote

Every variable, grouped by concern, is documented in .env.example. Copy it to .env and populate only what the operator use.

Docker Compose

The repo ships docker/mcp.compose.yml. It reads a sibling .env and publishes the HTTP server on :8000:

services:
  audio-transcriber-mcp:
    image: example/audio-transcriber@sha256:<digest>
    container_name: audio-transcriber-mcp
    hostname: audio-transcriber-mcp
    restart: always
    env_file:
      - ../.env
    environment:
      - PYTHONUNBUFFERED=1
      - HOST=0.0.0.0
      - PORT=8000
      - TRANSPORT=streamable-http
    ports:
      - "8000:8000"
    healthcheck:
      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
      interval: 30s
      timeout: 10s
      retries: 3

cp .env.example .env          # then edit WHISPER_MODEL and any other values
docker compose -f docker/mcp.compose.yml up -d
docker compose -f docker/mcp.compose.yml logs -f

Behind a Caddy reverse proxy

Expose the HTTP server on a hostname with automatic TLS. Add to the operator's Caddyfile:

# Internal (self-signed) — homelab .example.invalid zone
audio-transcriber.example.invalid {
    tls internal
    reverse_proxy audio-transcriber-mcp:8000
}

# Public — automatic Let's Encrypt
audio-transcriber.example.com {
    reverse_proxy audio-transcriber-mcp:8000
}

Reload Caddy:

docker compose -f services/caddy/compose.yml exec caddy caddy reload --config /etc/caddy/Caddyfile

DNS with Technitium

Point the hostname at the host running Caddy. Via the Technitium API:

curl -s "http://technitium.example.invalid:5380/api/zones/records/add" \
  --data-urlencode "token=$TECHNITIUM_DNS_TOKEN" \
  --data-urlencode "domain=audio-transcriber.example.invalid" \
  --data-urlencode "zone=arpa" \
  --data-urlencode "type=A" \
  --data-urlencode "ipAddress=192.0.2.10" \
  --data-urlencode "ttl=3600"

…or add an A record audio-transcriber.example.invalid → <caddy-host-ip> in the Technitium web console (http://technitium.example.invalid:5380). The ecosystem technitium-dns-mcp automates this as a tool.

Register with an MCP client

Add to the operator's client's mcp_config.json:

{
  "mcpServers": {
    "audio-transcriber": {
      "command": "uv",
      "args": ["run", "audio-transcriber-mcp"],
      "env": {
        "WHISPER_MODEL": "base",
        "TRANSCRIBE_DIRECTORY": "~/Downloads"
      }
    }
  }
}

For a remote HTTP server, point the client at http://audio-transcriber.example.invalid/mcp instead.