Configure your wake chain in 60 seconds
Explore freely—no account wall. Start with solo /wake-me on Telegram, Slack, or Discord,
and add teammates whenever you want backup.
Where should your agent wake you first?
Pick your primary channel (supports iOS, Android & Desktop).@Bot
Reply via 1-tap inline buttons, text, or a quick voice note—no bot token or chat ID to copy-paste.
Set your escalation chain & noise threshold
Never get woken at 3 AM forrm -rf dist.
Connect your AI client & fire a test wake
Test in browser right now, or mint your live MCP endpoint.claude mcp add --transport http wakeahuman https://wakeahuman.com/mcp
One wake, start to finish
Your agent filters out low-stakes noise, strips secrets before anything leaves your machine, and hands you—or your backup teammate—a 60-second decision brief.
-
Your agent
wake_human({ severity: "high", domain: "stripe-webhooks", question: "Run blocking SQL lock migration now, or ship KV dedup behind feature flag?" }) -
Secret scrubber
Redacts API keys, database URLs, and emails automatically. Suppresses trivial permission noise during quiet hours.
-
0s → 2m chain
Drops a silent Slack card on your desk at
0s. Still away at2m? Buzzes your Telegram with 1-tap options and a “Show Me” diff room link. -
You (or @alex)
“Go with Option B behind
ENABLE_WEBHOOK_GUARD. Queue the SQL migration for 9 AM.” -
Your agent
{ "status": "resolved", "decision": "option_b_kv_flag", "transcript": "Go with Option B behind ENABLE_WEBHOOK_GUARD. Queue the SQL migration for 9 AM.", "decided_by": "@you", "signed_receipt": "wah_sig_9f31c08ab42e" }
Built for builders running background agents
Not every agent block deserves an interruption. WakeAHuman separates routine prompts from decisions that actually need a human.
Threshold brain, not dumb relays
Separates trivial permission prompts from irreversible architecture, auth, billing, and prod schema decisions—with quiet hours built in.
Ack ≠ Resolve
Tap Ack (+15m) to pause the escalation timer while you inspect the diff on your phone so your teammate isn't woken unnecessarily.
Automatic secret & PII scrubbing
.env keys, JWTs, connection strings, and emails are masked into [REDACTED_*] tokens before a brief ever reaches Telegram, Slack, or Discord.
Deterministic timeout fallbacks
If nobody in your chain answers at 3 AM, your agent doesn't hang forever—it follows your policy to hibernate, open a safe Draft PR, or rollback.
Questions
How does it avoid waking me for trivial prompts?
You set your threshold filter and quiet hours in Step 2. Routine checks stay quiet, while high-stakes architecture, auth, billing, and production schema forks trigger your 0s → 2m → 5m escalation chain.
How do voice replies work?
Speak directly inside the mobile “Show Me” web room (or reply in Telegram), and your instructions return as structured JSON to your paused MCP session. Dedicated PSTN VIP phone calls & SMS arrive next.
Which AI coding clients are supported?
Any client supporting Streamable HTTP MCP servers: Claude Code, Cursor, Codex, Antigravity, Gemini CLI, Windsurf, Cline, and custom agent loops.
What happens if nobody in my chain answers?
You control the fallback in Step 2: hibernate pauses the run cleanly burning $0 tokens until morning, safe_default_draft_pr proceeds behind a feature flag in a Draft PR, or abort stops cleanly.