ctrlnode.ai

Remote control for your local AI agents — Claude, Copilot, Gemini, Cursor, Codex & OpenClaw.

Orchestrate multi-agent workflows, run scheduled routines

[ctrlnode.ai @ctrlnodeai]

CTRL NODE v2.2 is live.

Two updates that change how you run agents on your own machine:

🧠 Hermes — a stateful provider with persistent memory. Follow-up tasks keep context from the last run. No re-explaining the project every time.

🗂️ Work Directory — point an agent at a repo or folder and it reads and edits files in place. Optional focus paths. No more copying output from an isolated folder into your codebase.

Bridge is open source. Seven native providers. New `ctrlnode --setup` wizard.

See the full release notes:
https://t.co/J9sIoKp0pH

[ctrlnode.ai @ctrlnodeai]

"I completely agree. Limits force you to be more selective, and that's a good thing.

The idea is precisely to give you freedom only if you have that proven boundary. You can start with manual runs to validate, and then move to unlimited + scheduling once you have a clear signal.

How do you manage it with your agents in Hyperliquid?"

[ctrlnode.ai @ctrlnodeai]

Claude Code Routines limits you to 5 runs/day.
CTRL NODE gives you unlimited — and your code never leaves your machine.
Write it once. Schedule it. Done.

[ctrlnode.ai @ctrlnodeai]

I just published Building an outbound-only WebSocket bridge for local AI agents https://t.co/qVrJnK1e86

[ctrlnode.ai @ctrlnodeai]

.
Unlimited Claude Routines. Local. Free.
------------------------------------------

Claude Code Routines limits you to 5 runs/day.

CTRL NODE gives you unlimited — and your code never leaves your machine.

Write it once. Schedule it. Done.

https://t.co/K0ltOxHAt5

[ctrlnode.ai @ctrlnodeai]

https://t.co/tEj4ChWhsa

[ctrlnode.ai @ctrlnodeai]

Get started in 3 steps
1 — Sign up
Create an account at https://t.co/1sZS5zkHXb. You'll get a Pairing Token from Settings → Bridge.

2 — Install the Bridge
Linux / macOS — one-liner installer

curl -fsSL https://t.co/pVA7Va7K54 | sh

3 — Run it

[ctrlnode.ai @ctrlnodeai]

5/ The result: full Mission Control — pipelines, queues, monitoring — without asking clients to open ports or hand over credentials.

Repo: https://t.co/2Qx7ls4REN

[ctrlnode.ai @ctrlnodeai]

4/ The bridge is implemented in TypeScript and packaged with Bun into small executables for Linux/macOS/Windows, so it’s easy to install and run as a service.

[ctrlnode.ai @ctrlnodeai]

Your AI agent doesn't need you to tell it what to do next.

It needs a work queue.

CtrlNode's Kanban is built for agents, not humans. Tasks move from backlog → inbox → done autonomously. Blocked tasks surface automatically.

https://t.co/1sZS5zkHXb

[ctrlnode.ai @ctrlnodeai]

Most AI agent setups look like: prompt in → output out → pray.

This is what a real workflow looks like.

Visual pipeline builder for AI agents. Drag, connect, deploy. Runs headless on your VPS 24/7.

https://t.co/1sZS5zkHXb

[ctrlnode.ai @ctrlnodeai]

3/ When you trigger a task in CtrlNode, the backend sends a task:run command over that WebSocket. mc-bridge receives it, calls the Claude API locally using the client's own Anthropic key, and streams results back.

[ctrlnode.ai @ctrlnodeai]

2/ mc-bridge runs on the client's VPS and opens an outbound WebSocket to CtrlNode's cloud backend.

Outbound only. No open ports. No firewall rules. The client's server calls home — not the other way around.

[ctrlnode.ai @ctrlnodeai]

How do you control an AI agent running on a client's server without touching their firewall?
That's the problem we solved with mc-bridge. Here's how it works

1/ The typical approach: open a port, expose an API, hope nothing breaks. Bad for security. We hate it.

[ctrlnode.ai @ctrlnodeai]

CtrlNode is Mission Control for AI agents.

Visual pipeline builder. Kanban task queue. Outbound-only bridge so your agents run headless on your own VPS — no open ports, no vendor lock-in.

Following along 👇

https://t.co/1sZS5zkHXb