remoteagent.online  the open-source control plane for secure, cost-controlled AI agents

Open-source AI agent control plane: inspect it, fork it, self-host it

remoteagent.online is an open-source (MIT) control plane for AI agents. Read the code, verify what leaves your machine, and self-host the whole stack.

Updated 2026-08-22Reading time 5 min

Why open source matters for agents

An agent that runs commands on your machine should not be a black box. Open source means you can read exactly what it sends, where, and when. If you do not like a default, you change it.

What is in the repo

  • The agent daemon — runs on macOS, Linux, WSL2, and Raspberry Pi.
  • The control plane — dashboard, vault, budgets, audit trail.
  • The sandbox — allowlisted shell, workspace boundaries, deletes to trash.
  • Docs, examples, and a security model you can actually read.

Run it yourself

  1. Clone the repo: git clone https://github.com/MONAEXPERT/agent
  2. Install dependencies and start the daemon: remoteagent start.
  3. Point it at your own instance or use the hosted console.
  4. Bring your own keys, or run Ollama with no keys at all.

Hosted vs self-hosted

The hosted console is the fastest way to start — it manages the vault, budgets, and audit trail for you. Self-hosting keeps every byte on your own hardware and is fully supported by the same codebase.

License and lock-in

The project is MIT licensed. There is no per-seat lock-in, no proprietary format for your data, and no vendor that disappears with your audit trail — you can export it as JSONL.