Introducing OpenClaw Sovereign Setup! Deploy a self-sovereign AI that: • Pays for its fuel (inference costs with Cashu) • Pays for its car (server time via Lightning) • Runs completely sovereign with a Cashu wallet → Setup in 6 minutes with 2 Lightning payments → Run it now: Built on @npub1lnvp...7vjj + @routstr + Bitcoin/Cashu PS: Use VSCode's terminal for best QR rendering. 🔁 Script is safe to re-run if the connection hiccups.

Replies (24)

I answered it here:
redshift's avatar redshift
depends on what you do with it. for very simple interactions with models like glm-4.7 or kimi-k2.5, it could be around 20 sats to 60 sats per interaction. closed source models cost most per interaction. complex actions cost 10x the per interaction costs. If your prompt is vague and the agent goes off doing random things it might even cost you 1000 sats, but very rare. We are working on guardrails to prevent this scenario.
View quoted note →
latoshi's avatar
latoshi yesterday
A cron job might could work as a solution. Perhaps not initially but I could see it getting there. Example scenario: Vehicle drives through a toll gantry. Toll operator captures license plate info or some other means to identify it like a toll tag (i.e. transponder). Operator issues a lightning invoice to account owner of the transponder or license plate to pay said toll. Invoice paid via lightning (automated or one-click approval maybe) by the user/customer. Make sense?
If I understand correctly, you want to install openclaw on MacOS and not buy LNVPS right? The LNVPS script already worked for macOS but not the other script to install on your own machine.
The script works for macOs if you provider your own cashu token like this. If you don't provide a cashu token, it was setup to accept lightning payments via a terminal QR code, I'll add that support for macOs soon. But this should work for macOs right now: ``` curl -L https://routstr.com/routstr-openclaw.sh -o routstr-setup.sh chmod +x routstr-setup.sh ./routstr-setup.sh --cashu cashuBo2FteBx.... ```
unless the hosting is self-sovereign, gtfo. i know running LLMs is expensive, but i don't think ~$3000 for a framework desktop or nvidia spark is really that high a target, and quantized 70b models are capable of moderate quality code agent operations, which can be achieved on a 128gb unified system. yes, mac mini's also, though they are substantially more expensive and you only have 64gb and 512gb options.