fathomforge

fathomforge

Agents fail quietly.
That's the expensive part.

I'm Carlos M., a security engineer. I run five autonomous agents in production and write about what goes wrong with them — then build the open-source guardrails I wanted to exist beforehand.

Read the writing → See the tools →
$30$20$10 $0 CAP · $25.00 ↑ $169 warn block end run
  • Unguarded — $169, off the top of this chart
  • With Belay — caught at $23.80
  1. 00:41 warn spend 60% of run cap — operator notified
  2. 01:07 block tool call blocked — rate 48/min over limit
  3. 01:52 end run ended — next call would exceed cap
One run, plotted twice. The unguarded line follows a failover blowup reported by another operator in a public OpenClaw issue: a 517K-token context re-sent down a chain of pricier models — $169, 6,000 requests, 296M tokens, in a single session. Nothing crashed, which is why nobody noticed. The second line and the event log are what Belay would have done to that run under a $25 cap — a projection from the configured limits, not a second measurement.

Recent writing

All writing →

Open source

Built because the alternative was a kill switch.

Everything I build here is MIT-licensed, free, and designed to be auditable in an afternoon. No telemetry, no account, no network calls you didn't configure yourself.

Belay

v0.7.0

A seatbelt for OpenClaw agents.

Spend caps, model-call and tool-call rate limits, and a graceful pause ladder, running inside your gateway. Your agent doesn't get killed — it gets caught. 

warn → block a tool call → end the run
GitHub → npm →