Belong.net Logo
Belong AI Deploy

Open Source

Self-host aideploy from the public repository and understand the boundary between the public stack and managed platform.

Open Source

The public aideploy source is at github.com/belongnet/aideploy.

It is the deployment and operations foundation for private OpenClaw and Hermes Agent installations: lifecycle CLI, runtime assets, OpenTofu modules, readiness proof, backups, upgrades, diagnostics, teardown, and public contracts.

View the public repository

Public self-host or hosted platform?

Public self-host beta

Choose this when source access and local lifecycle control matter most.

  • Setup: tagged source checkout and local CLI
  • Cloud: DigitalOcean golden path
  • Runtime: OpenClaw or Hermes Agent
  • AI: OpenAI, Anthropic, or Kimi API key
  • Messaging: Telegram
  • Scale: one deployment at a time
  • Operations: local status, VM backup, blue/green upgrade, and teardown
  • Network: Tailscale-private browser and runtime

Hosted + managed platform

Choose this for guided setup, teams, agencies, fleets, and managed operations.

  • Setup: guided web wizard and API
  • Cloud: DigitalOcean, OVHcloud, Google Cloud, AWS, and Scaleway
  • Runtime: OpenClaw or Hermes Agent
  • AI: ChatGPT and Claude connection flows plus OpenAI, Anthropic, Gemini, Kimi, and DeepSeek keys
  • Messaging: eight OpenClaw channels; Telegram and WhatsApp beta bridge for Hermes
  • Scale: 1–16 agents, isolated workspaces, and virtual-office presets
  • Operations: Command Center, monitoring, offsite backup, managed updates, REST API, and MCP
  • Network: the same private data plane, plus signed webhook ingress only when required

Microsoft Azure is not a new-deployment path in either column. Public Azure files are retained as validation/reference contracts, and hosted Azure code is retained only for compatibility with existing resources.

What the public beta proves

The public golden path is DigitalOcean + Telegram, using either OpenClaw or Hermes Agent and an OpenAI, Anthropic, or Kimi API key.

A successful up does more than create a VM. It waits for the selected runtime, AI path, Telegram path, and private HTTPS browser surface to become usable. It also:

  • keeps runtime ports on loopback and exposes them through Tailscale HTTPS
  • limits Telegram access to the numeric owner ID you provide
  • pins runtime inputs and verifies downloaded OpenTofu checksums
  • tags resources so doctor can find orphans even when local state is damaged
  • stores local access material with mode 0600
  • performs deterministic teardown and scrubs secrets after success

Follow the repository's self-host guide for the current release command and prerequisites.

Releases and supply-chain verification

Public releases include deterministic source archives, manifests, SHA-256 values, and keyless Sigstore bundles. Runtime images are published by digest and signed from the repository's release workflow.

Always use the verification command shipped with the release you deploy; do not copy a digest from an older guide.

Licensing

The deployed stack, OpenTofu modules, and contracts are Apache-2.0. The cli/ directory uses FSL-1.1-ALv2: it is free to use, modify, and self-host, while competing managed-service use is restricted for two years per release before that release converts to Apache-2.0.

Read the repository's licensing section and cli/LICENSE.md before redistribution.

Build with us

Contributions that shorten the path from clone to a real agent reply, strengthen safety, or graduate another provider into the live CLI are welcome.

For the fastest managed path, return to the Quick Start. For the hosted cloud matrix, see Cloud Providers.

Copyright © 2026