What is the OpenClaw safeguard compaction billing bug?

By default, OpenClaw uses compaction.mode=safeguard, which calls the LLM API every 30 minutes on every active session including idle ones. Each call sends ~33K tokens to the LLM and discards the response, generating approximately $38/month in silent background billing per deployment.

How do I fix the OpenClaw compaction idle billing bug?

Set compaction.mode to a different value in your OpenClaw configuration, or disable automatic compaction entirely for deployments that use long-running idle sessions. Track GitHub issue #34935 on the openclaw/openclaw repository for an official fix.

How many API calls does OpenClaw make on idle sessions?

Approximately 48 unnecessary LLM API calls per day per idle session with the default safeguard compaction mode. Each call sends the full session system prompt (~33K tokens) before correctly cancelling the compaction.

Is the OpenClaw compaction billing bug fixed?

As of March 2026, the issue is tracked as GitHub issue #34935 and has not been officially patched. The workaround is to manually change the compaction mode in your OpenClaw settings.

← Back to dashboard

clawsmith.com/signal/openclaw-safeguard-compaction-idle-billing-bug

⚠ IssueUnderservedLive

OpenClaw Safeguard Compaction Makes 48 Unnecessary LLM API Calls/Day on Idle Sessions

Default compaction.mode=safeguard calls the LLM API every 30 min on all active sessions including idle ones. Each call sends ~33K tokens and is discarded, generating ~$38/month in silent billing. GitHub issue #34935.

Product Idea from this Signal

A CLI dashboard that calculates real-time OpenClaw token spend across providers and finds the cheapest model path before the bill arrives

5.1k ▲

Anthropic just blocked flat-rate Claude subscriptions for OpenClaw users starting April 4, 2026. Developers who were paying $20/month are now facing $200-623/month API bills with no warning. The built-in /status command only shows the last response cost, not cumulative spend, and there is no way to compare what you would pay across providers. This tool reads your gateway logs, calculates actual token spend broken down by model, skill, and conversation, then shows what each session would cost under every billing option so you can switch before the surprise bill hits.

CLIOPEN-SOURCEDEVTOOLCOST-OPTIMIZATION

Competitive35 leadsView Opportunity →