Agent configs drift. Now you can see exactly where.

Scan your OpenClaw workspace. See config drift, find truncation risks, track costs — all in your browser.

✦ Interactive Demo

◈

Driftwatch/Agent MapDemo

A health check for your agent workspace.

The overview aggregates findings from every module — file sizes, config review issues, drift status, and cost signals. The health score is a weighted composite: critical findings count more than warnings. Click any card to jump to the detailed view.

Bub's Agent Architecture

5 agents

Total Files

Agents

Memory Entries

Skills

Agent FleetView all →

Bubclaude-opus-4-6

Sonnetclaude-sonnet-4-6

Coderdeepseek-chat

Analystdeepseek-chat

Localdeepseek-chat

Config ReviewView all →

AGENTS.md

AGENTS.md is 20,110 characters — 14,990 over the 5,120-character critical limit. That's ~3,748 extra tokens injected every message.

(all bootstrap files)

Total bootstrap size is 21,868 characters — 6,508 over the 20,480-character critical limit. That's ~5,467 tokens injected per message.

AGENTS.md

AGENTS.md is 20,110 characters — OpenClaw truncates it to 20,000. Characters 14,001–16,110 (2,110 chars) are invisible to your agent. The head (70%) and tail (20%) are kept; the middle disappears.

Bootstrap Budget

31,778 / 20,000 chars159%

AGENTS.md20,146

TOOLS.md4,500

SOUL.md3,500

File HealthView all →

SOUL.md✓

AGENTS.md✓

MEMORY.md✓

TOOLS.md✓

HEARTBEAT.md✓

Memory

10 files✓ 10 entries

Files

Agents

Memory

Skills

Score

Agent Fleet

BubOpus⚠

SonnetSonnet✓

CoderDeepSeek✓

AnalystDeepSeek✓

LocalDeepSeek⚠

Workspace Files

SOUL.mdAGENTS.mdMEMORY.mdTOOLS.mdHEARTBEAT.mdUSER.mdIDENTITY.mdBOOTSTRAP.md10 memory files

Health Score10/10

OpenClaw cuts the middle of oversized files — your agent sees a marker, but you get no warning about what was lost. We show you exactly which lines your agent can’t see.

Built for the moment you realize something’s off

Your agent forgets things

OpenClaw config files control what your agent knows. When they get too long, the middle gets cut — silently. Your agent loses instructions you wrote and you get no warning.

Changes pile up invisibly

You edit AGENTS.md, tweak a heartbeat, add a skill. Six weeks later you can’t remember what changed. Drift compounds. Bugs hide in the gap between what you wrote and what’s actually loaded.

Driftwatch makes it visible

Scan your workspace. See which files are oversized, which sections get truncated, how costs are trending, and what changed since last time. Fix problems before they cost you.

That was a demo. Now see yours.

Scan Your Workspace →View on GitHub

What’s available now

🗺 Architecture map🔍 Config review📊 Drift detection💰 Cost tracking✏️ In-browser editor

Scan → Review → Fix → Track

Coming Soon — Driftwatch Pro

More is coming

Driftwatch Pro will show you what your config is actually costing you — and how to fix it.

Why does my agent forget instructions?See exactly which rules survive long conversations and which ones vanish silently.
Where are my tokens going?Your config files get re-injected every message. We’ll show you the per-message cost and where to cut.
Why do my subagents behave differently?Three of your eight config files are invisible to subagents. We’ll show you which ones.
Which files are starving which?OpenClaw loads your files in a fixed order. The last file gets whatever budget is left. We’ll show the cascade.

No spam. Unsubscribe anytime. We'll only reach out when something real ships.