Best practices for using ACP as a remote node to connect to my OpenClaw gateway?
Hi folks, I’m trying to set up a remote workflow and want to confirm the recommended pattern for ACP.
- Remote compute host (Vast): where I want to run ACP/Codex workloads
- Gateway is kept on
- I connect Vast -> Claw via SSH tunnel
Then I start ACP:
I can see logs like:
-
-
But:
-
- If I bind to a Discord session key (e.g.
2. What is the correct way to create/use an ACP-enabled session (vs normal Discord session)?
3. Should I always use
4. Any recommended production pattern for persistent ACP on remote hosts (tmux/systemd/autoreconnect)?
Would appreciate a minimal “known-good” flow. Thanks!
My setup
- Gateway host (Claw): runs OpenClaw gateway- Remote compute host (Vast): where I want to run ACP/Codex workloads
- Gateway is kept on
bind: loopback- I connect Vast -> Claw via SSH tunnel
What I’m doing
On Vast, I create a tunnel:Then I start ACP:
I can see logs like:
-
[acp] ready-
[acp] gateway reconnectedBut:
-
openclaw nodes listPaired: 0- If I bind to a Discord session key (e.g.
agent:main:discord:channel:...ACP_SESSION_INIT_FAILED: Session is not ACP-enabledQuestions
1. Is it expected thatnodes listopenclaw acp2. What is the correct way to create/use an ACP-enabled session (vs normal Discord session)?
3. Should I always use
/acp spawnopenclaw acp --session <key>4. Any recommended production pattern for persistent ACP on remote hosts (tmux/systemd/autoreconnect)?
Would appreciate a minimal “known-good” flow. Thanks!
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
137,667Members
Resources
Similar Threads
Was this page helpful?
