Keep Msty Studio.
Give your models your machines.
AI Commander is not an alternative to Msty Studio, Claude Code, Codex, ChatGPT, or any other AI client. It is the remote-machine harness those tools plug into over MCP: outbound-only shell, detached jobs that outlive the chat, and file transfer on hardware you already own.
- Install the host agent on the machine you want your AI to control. It shows a private machine code.
- Connect your AI: paste this prompt, then replace
AIC-…with your code —Help me connect this AI client to AI Commander using https://aicommander.dev/start.md. Then check the status of my machine: AIC-…— or add the connector URLhttps://aicommander.dev/mcpand sign in with an email code (per-client steps).
Sibling harness pages: Jan · Chatbox · LM Studio · AnythingLLM · TypingMind · Open WebUI · Claude Code · Codex · Claude Desktop — and the rest on Connect your AI client.
Connect Msty Studio (Toolbox, Streamable HTTP)
Msty Studio implements MCP in its Toolbox: local servers over STDIO / JSON and remote servers over Streamable HTTP (Msty Studio docs: Toolbox tools). Add AI Commander’s hosted endpoint as a remote tool:
- Open the Toolbox and click Add New Tool.
- Choose HTTP and enter the MCP server URL
https://aicommander.dev/mcp. - Add the header
Authorization: Bearer <your-api-key>(or see below for the no-account URL), give the tool a name and save it. - Add the tool to a toolset and select that toolset in your chat (Msty Studio docs: Toolsets).
URL: https://aicommander.dev/mcp Header: Authorization: Bearer <your-api-key>
Create an API key in the AI Commander dashboard (see Accounts & API keys) and send it as the Authorization: Bearer header so the agent can list your machines and reach them by alias.
No account yet? Use https://aicommander.dev/mcp?anonymous=1 as the URL instead (no header) for a code-only first hour, reaching machines by their AIC-… session code.
Prefer a local process? Choose STDIO / JSON instead and paste the stdio bridge (needs Node.js 20 or later; drop the env block if you have no key — the bridge then connects anonymously):
{
"command": "npx",
"args": ["-y", "@aicommander/mcp"],
"env": {
"AICOMMANDER_TOKEN": "<your-api-key>"
}
}
Get the details right
- Transport: Msty Studio connects to remote servers over Streamable HTTP (not SSE), which is what
https://aicommander.dev/mcpserves. - Long work: start detached jobs (
remote_job_start, thenremote_job_status/remote_job_logs) instead of one long tool call. - Models: pick a model with solid tool calling; the tools are the same whichever model you chat with.
By default, API keys and OAuth connectors only act while the account owner has opened the AI Commander dashboard within its access window (72 hours by default; 24 h, 48 h or 7 days in the dashboard → Security tab; the gate can also be disabled) — see Connect your AI client. AI Commander’s endpoint is first-party and only acts on machines where you installed the agent yourself. Menus and options can differ between Msty Studio versions — check the linked Msty Studio docs for your build.
What you get that the chat alone cannot
- Remote shell on Linux, macOS, and Windows — by session code or saved alias.
- Detached jobs for builds, test suites, and training on the box that has the repo.
- Fleet list + GPU facts so your agent picks the right machine before it burns an hour.
- File courier (Pro) for checkpoints and diffs up to 100 MiB.
- Outbound-only agent on each machine — no inbound ports, no exposed SSH, no VPN to set up.
Not a replacement — a multiplier
| You already use | AI Commander adds |
|---|---|
| Msty Studio | Toolbox → Add New Tool → HTTP → https://aicommander.dev/mcp |
| Claude Code | claude mcp add --transport http … → remote exec + jobs |
| Codex / ChatGPT | Same MCP endpoint or REST /api/v1/exec |
How Msty handles MCP servers
- MCP in Msty Studio lives in the Toolbox. Local servers are added as STDIO/JSON, and remote ones over Streamable HTTP.
- SSE servers are not supported. A server that only offers SSE has to go through a bridge such as
mcp-remote. AI Commander's endpoint is Streamable HTTP, so it connects directly. - Studio Web can't start local tools on its own. It needs Studio Desktop reached through a Remote Connection or the Sidecar.
- STDIO configs can use variables like
{var:Label:default}, which Msty asks for when the tool is added. - The Tool Console lets you call a tool by hand before a model ever sees it.
- A tool must be part of a toolset, and the toolset enabled in the chat, before a model can call it.
Checked against Msty docs: Toolbox tools in November 2026. Client behaviour changes between releases; the official docs win if they disagree.
A workflow that suits Msty
Msty is popular for running local models. Pairing it with AI Commander keeps the model on your desk and the commands on the machines that need them.
- Add
aicommanderto the Toolbox as a Streamable HTTP tool with yourAuthorization: BearerAPI key header. - Open the Tool Console and call
list_machinesonce, to confirm the connection before any model is involved. - Put AI Commander in a small "Ops" toolset, enable it in a chat with a tool-capable local model, and ask for routine checks such as service status or log tails through
remote_exec.
Limits and gotchas in Msty
- For the stdio bridge (
npx), Msty recommends installing Node through Volta rather than NVM. - Adding the tool isn't enough: if it isn't in the active toolset, the model won't see it.
- When a tool fails to start, the logs are under Sidecar → View Logs.
FAQ
https://aicommander.dev/mcp endpoint with an API key header and skip Node entirely.Next steps
Install the agent on a machine you control, add AI Commander to the Msty Toolbox as an HTTP tool (https://aicommander.dev/mcp), then ask your model for a real task on a remote box.