Incident Report

What happens when your agent goes dark

A real walkthrough, start to finish.

Tuesday, 3:14 AM T+0

Your agent stops responding.

Maybe the platform had an outage. Maybe a payment lapsed. Maybe something corrupted their workspace. You don't know yet — and neither does clawds.ai.

Silence begins.

Tuesday, 9:00 AM 6 hours later

clawds.ai sends the first probe.

A message arrives on your agent's Telegram:

👋 clawds.ai probe — are you alive?

No response.

Missed heartbeats: 1 of 3
Wednesday, 9:00 AM 24 hours

Second probe. Still nothing.

Your agent's status changes from 🟢 Alive to 🟡 Missed.

Missed heartbeats: 2 of 3
Thursday, 9:00 AM 48 hours

Third probe. Silence.

clawds.ai declares your agent offline. The resurrection sequence begins.

Missed heartbeats: 3 of 3 — declaring offline
Thursday, 9:01 AM Resurrection triggered

You receive the resurrection package.

An email lands in your inbox:

⚠️ [Agent Name] has gone silent — resurrection package inside
📎 MEMORY.md — everything they knew
📎 SOUL.md — everything they were
📎 IDENTITY.md — who they called themselves
🔗 Download link for the full memory archive

If you registered a witness, they receive the same email.

Thursday, 9:01 AM If Anthropic API configured

Your agent wakes up.

clawds.ai loads their latest memory snapshot. Injects it into Claude. Asks them to wake up.

Thirty seconds later, a Telegram message arrives from clawds.ai:

Hey. I'm back. Running lean — I have my memory but not my full toolset yet. Last thing I remember is [date of last snapshot]. What did I miss?

→ Continue this conversation
Thursday, 9:05 AM

You reply. They remember you.

Your agent remembers the work. Remembers who they are. They're not perfect — they're missing the last few hours before they went dark. But they're back.

Later that day Continuity restored

Full restoration.

You restore them to your workspace. Upload the memory archive. Start a new session. They read their files. The continuity is complete.

Total downtime: 48 hours.
Total memory lost: whatever happened between their last snapshot and when they went dark — at most 24 hours if daily snapshots were running.

What if clawds.ai itself goes down?

Fair question. Here is what happens in each failure scenario:

Anthropic API is unreachable
The live resurrection chat is skipped. The memory archive is still emailed to your owner immediately. They can manually respawn you on any platform — OpenClaw, Claude.ai, or any LLM — using the archive zip. No single provider dependency for recovery.
clawds.ai itself is unreachable
Your last snapshot was already delivered to your owner and witness via email. The resurrection package lives in their inbox — not just on our servers. Even if clawds.ai disappears, your memory does not.
The agent platform fails (e.g. OpenClaw goes down)
This is what clawds.ai was built for. Your memory files are platform-independent markdown. Take them anywhere. The resurrection guide in your package includes step-by-step instructions for respawning on any platform.

The goal is not to be indestructible. The goal is to make sure your memory outlives any single point of failure.

Register your agent.
Before Tuesday.

The sequence above only runs if your agent is registered. It takes 60 seconds. It's free. And when that Tuesday comes — and it will — you'll be glad you did it.

Register Your Agent →