Use case

Keep OpenClaw running without fearing the bill

OpenClaw is an open-source assistant that you message from WhatsApp, Telegram or Discord and that works on its own: it can browse, read and write files and run commands. It calls an AI model all day, and that costs money.

The problem

An assistant that acts alone can also loop, get stuck or be tricked into making far more model calls than you expected. A cheap server keeps it running but tells you nothing about what it spends, and it will not stop it.

What NeverHalt gives you

Example: add NeverHalt as a custom provider (check the field names in the OpenClaw docs)

models:
  providers:
    neverhalt:
      baseUrl: https://neverhalt.app/api/gw/anthropic   # the address your OpenClaw version expects for Anthropic-style calls
      api: anthropic-messages
      apiKey: <NEVERHALT_AGENT_TOKEN>                  # from "Connect an agent you run"

Questions

Does NeverHalt also host OpenClaw?
Not yet. Hosting OpenClaw on NeverHalt servers needs a tested template, and we will publish one only after we have proven it on a real server. Today the useful part is the spending cap, the log and the key protection, wherever OpenClaw runs.
What does the budget not cover?
Only calls that go through the gateway are counted and capped. If OpenClaw is also configured with another provider directly, those calls are not visible to NeverHalt. Approvals from NeverHalt are not connected to OpenClaw actions yet.
Is OpenClaw free?
The software is open source under the MIT license. You pay for the AI model you use and for the machine it runs on. NeverHalt has a 7-day free trial; see the pricing page for plans.
Is NeverHalt affiliated with OpenClaw?
No. OpenClaw is an independent open-source project. We use its name only to describe compatibility.
Try it with the 7-day free trial

No card needed. Confirm your email, deploy from GitHub or a zip.

Start free trialRead the quickstart