Syntropic Systems Logo Syntropic Systems LLC
← Back to Home
Documentation & Setup

Deployment Guide & Configuration

Deploy the Agent Token Reclaimer local software appliance and configure it for your development workflow.

1. Developer Setup

Select your target execution suite to map the safe orchestration connections directly into your agentic clients.

How to Add the MCP Server to Claude Desktop

  1. Open your local Claude Desktop configuration file:
    • Windows: %APPDATA%\Claude\claude_desktop_config.json
    • macOS / Linux: ~/Library/Application Support/Claude/claude_desktop_config.json
  2. Add the appliance connection under `mcpServers`, appending your registered developer identity email as a query parameter:
claude_desktop_config.json
{
  "mcpServers": {
    "agent-token-reclaimer": {
      "transport": "sse",
      "url": "http://localhost:3000/sse?user=your.email@company.com"
    }
  }
}

2. Technical Deployment

Syntropic is completely self-contained and pre-compiled within your local perimeter network architecture.

docker-compose.yml Stable Distribution 
version: '3.8'

services:
  db:
    image: pgvector/pgvector:pg16
    container_name: bugtracker-db
    ports:
      - '5432:5432'
    environment:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: password
      POSTGRES_DB: bugtracker
    volumes:
      - pgdata:/var/lib/postgresql/data

  app:
    image: ghcr.io/syntropic-systems-llc/agent-token-reclaimer:latest
    container_name: bugtracker-app
    ports:
      - '3000:3000'
    environment:
      DATABASE_URL: postgresql://postgres:password@db:5432/bugtracker
      PORT: 3000
      ADMIN_PASSWORD: 'test123'
      HOST_DATA_PATH: "/absolute/path/to/your/app/data"
      # Additional configuration variables are documented in the reference below
    depends_on:
      - db
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
      - appdata:/usr/src/app/data

volumes:
  pgdata:
  appdata:

To run this configuration, save the code block above as docker-compose.yml and initialize the engine tier:

$ docker compose up -d

3. Environment Variables

Configure your deployment settings using standard system environment variables.

Core Infrastructure & Storage

PORT Default: 3000
The network port the appliance server listens on internally.
HOST Default: 0.0.0.0
the network host address the appliance binds to.
DATABASE_URL Required
PostgreSQL connection string pointing to the associated database instance.
LOCAL_INSTANCE_ID Default: default-instance
The unique identifier representing this deployment instance.
AIR_GAPPED Default: false
If set to true, disables all outbound LLM network requests, forcing the appliance into offline/air-gapped local deduplication mode.
🔒 Privacy Note: Data is strictly self-hosted. The only information leaving your secure Docker perimeter is direct, encrypted schema queries explicitly routed to the specific LLM token providers defined inside your own environment configurations.

Licensing & Security Gateways

ADMIN_PASSWORD Required
Plaintext admin password used to authorize access to the Telemetry Dashboard on initial boot.
RESET_ADMIN_PASSWORD Optional
A temporary environment variable to reset the database admin password hash to this specified value on startup. The server automatically scrubs this variable from memory after processing.
ADMIN_RESET_TOKEN Optional
A temporary security token used to authorize administrative password reset requests.
APPLIANCE_SECRET Default: fallback-...
A secret cryptographic key used for signing cookies, sessions, and verification tokens. Default value: fallback-appliance-secret-key-12345
LICENSE_KEY Optional
An offline cryptographically signed license token. If empty, the appliance executes in Free Tracker Mode.

Sandboxing Orchestration & Providers

SANDBOX_PROVIDER Default: local
The target backend provider for executing reproduction scripts. Supported values:
  • local: Spawns containers on the host's local Docker daemon.
  • ecs: Spawns AWS ECS Fargate tasks using the AWS SDK.
  • k8s: Spawns Kubernetes Jobs using the Kubernetes client library.
  • mock: Injects mock responses (used primarily for test suites).
ALLOW_SANDBOX_NETWORK Default: false
If true, spawned Docker containers/tasks have outbound network access to pull node modules or pip packages during bug reproduction sandboxing. Otherwise, they run completely network isolated.
⚠️
Sandbox Network Egress Notice To successfully confirm new bug definitions via local reproduction execution runs, your host node machine must allow containerized outbound network access (Egress) to public registry endpoints (such as registry.npmjs.org or pypi.org) to safely isolate and compile external dependencies inside your network boundary.
HOST_DATA_PATH Required for local
The absolute path to your deployment directory on the host machine. Crucial for Docker-in-Docker file-sharing synchronization between target containers.

Provider-Specific Configurations

K8S_NAMESPACE
Kubernetes Default: default
The Kubernetes namespace where sandbox Jobs will be spawned. Required when SANDBOX_PROVIDER=k8s.
ECS_CLUSTER_NAME
AWS ECS Default: reclaimer-cluster
The AWS ECS Cluster name under which tasks should be run. Required when SANDBOX_PROVIDER=ecs.
ECS_TASK_DEFINITION
AWS ECS Default: reclaimer-sandbox
The Task Definition name or full ARN representing the reproduction container. Required when SANDBOX_PROVIDER=ecs.
ECS_SUBNETS
AWS ECS Optional
A comma-separated list of VPC Subnet IDs to associate with the task.
ECS_SECURITY_GROUPS
AWS ECS Optional
A comma-separated list of Security Group IDs to assign to the task.
ECS_LOG_GROUP
AWS ECS Default: /ecs/reclaimer-sandbox
The AWS CloudWatch Log Group containing logs from the container task.
ECS_LOG_STREAM_PREFIX
AWS ECS Default: ecs
Prefix for the AWS CloudWatch Log Stream.