HybridClaw v0.41.0
Your agent writes a daily brief at the time you pick
HybridClaw 0.41.0 lets your agent write you a daily brief at a time you choose. Tracked work now survives restarts, and a new browser tool runs JavaScript in the page. Some older compatibility paths are retired too, so read the upgrade steps first.
Daily briefs
One brief a day, in your time zone
Opt-in. Decisions first, then prepared work, then what's coming up.
# save the schedule /schedule routine --json daily-brief <base64url-JSON> { "revision": null, "enabled": true, "hour": 7, "minute": 30, "weekdays": true, "tz": "Europe/Berlin" }
Daily briefs are opt-in. Each one starts with what needs a decision today, then prepared work, then upcoming commitments. A compatible app saves the hour, minute, weekdays and time zone with one schedule command. You get one edition a day, even across restarts. A run over five minutes late counts as missed, and retries reuse the saved brief.
Follow-through
Tracked work keeps its state through restarts
Tracked items now keep their evidence, check-ins and draft reviews through restarts, and every device reads the same record. An item is in progress, needs you, waiting, or blocked. It reaches completed only when you confirm the outcome. The overview screen arrives in compatible mobile apps.
Draft reviews
Approving a draft records your decision. It sends nothing.
Email drafts get one shared review that you can edit, defer, approve, decline or reopen. Every change is checked against the latest revision. Approving a draft records your decision and sends nothing. Sending stays in the app's own send flow.
browser_evaluate
browser_evaluate reads a page in one call
Runs JavaScript in the open page and returns JSON.
// every h3 heading on the rendered page [...document.querySelectorAll("h3")] .map(h => h.textContent.trim()) // promises are awaited: the page's own API fetch('/api/items').then(r => r.json())
The new browser_evaluate tool runs a JavaScript expression in the open page and returns the value as JSON. Promises are awaited, so it can fetch the page's own API with its cookies. Inline results stop at 48,000 characters. Pass a path and the whole result goes to a workspace file, even a CSV.
Where it stops
The script runs as the page, so it has guards
Because the script runs as the page, it has guards. Private hosts are refused unless you allow them. It stays off while a secret the agent entered is still on the page. Checkout and cancellation pages need your approval. And the mac-cua provider doesn't support it.
Telemetry without a bot
Telemetry works without a chatbot
Audit ingestion now uses the installation's node ID and your HybridAI API key.
"observability": {
"baseUrl": "https://hybridai.one",
"botId": "",
"label": "",
"environment": "prod"
}
Audit ingestion used to wait for a chatbot ID. Now it uses the installation's node ID and your HybridAI API key. Remove the bot ID setting from your configuration, and use an endpoint that accepts a node ID without one.
Removed compatibility paths
Retired compatibility paths, and what replaces each one
/api/plugin-webhooks/twilio-voice/webhook
Several compatibility paths are retired. Move old browser settings into the provider's plugin entry and enable it. Enable the Twilio voice plugin and point your number at its webhook before you upgrade. Harness evolution runs from a source checkout. And schema 79 drops the old Teams user table.
How to upgrade
Back up, then upgrade
Daily-brief controls and the follow-through overview need compatible app versions. Briefs stay off until you opt in.
WhatsApp's 0.1.x plugin stays supported until v0.42.
github.com/HybridAIOne/hybridclaw
/releases/tag/v0.41.0
Back up your configuration and database first. With pre-plugin browser or Twilio settings, or a Teams database older than schema 73, start 0.40.8 once before moving on. Then install 0.41.0. Rolling back to 0.39 needs that backup. The daily brief at the time you pick also needs a compatible app, and it stays off until you opt in.








