Set up Toggle for
the way you work.

Toggle can do a lot. This page walks you through each feature so you can turn on exactly what you want — from basic recall to full autopilot.

Get started

1

Get your API key

Sign up at x.toggle.pro, install the Chrome extension, then generate your key from the integration page.

2

Install the Toggle Skill

Make sure you have the ClawHub CLI installed. If not, run:

Terminal
npm i clawhub

Then install the Toggle skill:

Terminal
clawhub install aleksandar-jive/toggle
3

Configure & launch

Paste this into your Claw. It sets your API key, enables auto-sync every 3 hours, and turns on all default features.

You: Set up Toggle with my API key: [paste key here]. Enable auto-sync every 3 hours. Turn on daily digest at 6 PM and stale project nudges. Persist all data to memory. If I haven't stated my timezone also ask me about it before proceeding.

That's it. Your agent now has full context of your browser activity. Everything below is optional tuning.

Choose what to turn on.

Each feature builds on the last. Start with what feels right and add more as you get comfortable.

What it does
“What was I working on?” — gets an answer based on actual browser activity, not memory
“Pick up where I left off” — finds your last deep work session and describes exactly where you stopped
“What did I do yesterday?” — full breakdown by project, time, and focus score
Example conversation
You: Where did I leave off on the API project?

Claw: Tuesday afternoon (2:15–4:30 PM) you were deep in the Kalshi API docs — 1.5h across 3 sessions. Last session ended on the authentication endpoint. Focus: 87. Want me to pull up those docs?

No setup needed beyond the API key. Just start asking.

What it does
Fetches your activity every hour and saves it to agent memory
Enables multi-day history — your agent remembers last week, not just today
Required for stale nudges, pattern detection, and predictions to work
Tell your Claw
You: Set up hourly auto-sync for Toggle
What you get every day
Total work time, session count, and deepest focus area
Open items you didn't finish — so nothing falls through the cracks
Day-over-day comparison — focus trends, context switches, time spent
What it looks like
Claw: Your day: 5.2h across 4 sessions. Deepest focus on auth migration (2.1h, 94 focus). Stripe webhooks still open — you were 80% through. Context switches down to 3 from 6 yesterday.
Tell your Claw
You: Set up a daily digest at 6 PM

You can change the time anytime. Just say “change my digest to 8 PM.”

Two types of nudges
Stale project — project has 2+ hours invested but no activity in 48h. Your agent pings you before it goes cold.
Focus alert — 3+ project switches in 30 minutes with focus score below 30. Your agent offers to help you lock in.
Stale nudge example
Claw: You haven't touched “Landing Page Deploy” in 3 days. Last session: 1.5h on responsive layout. Want me to pull up where you left off?

Activated automatically once auto-sync is running and history builds up. Say “I dropped that” to dismiss a project permanently.

How it works
Scans 7 days of history for workflows that repeat 3+ times
Matches project sequences — same tools visited in the same order
Proposes an automation pipeline — only if the required skills are installed
Example
You: What repetitive stuff could you automate for me?

Claw: I noticed you check GitHub PR comments, update Notion, then post in Slack — 4 times this week, ~12 min each. Want me to automate that pipeline?

Requires auto-sync to be running. The longer it runs, the better the pattern detection gets.

Three types of predictions
Routine prediction — “It's Tuesday 10 AM — you usually start with PR reviews. Want me to pull up open PRs?”
Next-task prediction — “You just wrapped PR reviews. You usually hop into Notion next — want me to open your sprint page?”
Session endurance — “You're 85 minutes into the API integration — around when you usually take a break. Want me to bookmark where you are?”

Predictions activate automatically once 5 days of data exist. They self-correct — predictions you ignore get retired silently. Say “stop predicting” to pause for 7 days.

Not sure? Pick a preset.

Three ready-made configurations depending on how much you want your agent to do.

Most users

Daily Driver

Auto-sync + daily digest. Fresh context and a summary every evening.

RecallAuto-syncDigestNudgesPatternsPredictions
Send this to your Claw
Set up Toggle with hourly auto-sync and a daily digest at 6 PM. Persist all data to memory.
The power user

Full Awareness

Everything proactive — nudges, focus alerts, pattern detection.

RecallAuto-syncDigestNudgesPatternsPredictions
Send this to your Claw
Set up Toggle with hourly auto-sync, daily digest at 6 PM, stale project nudges after 48h, and focus alerts when my score drops below 30. Persist all data to memory.
The believer

Full Autopilot

Everything on. Your agent learns your routines and anticipates what's next.

RecallAuto-syncDigestNudgesPatternsPredictions
Send this to your Claw
Set up Toggle with everything enabled: hourly auto-sync, daily digest at 6 PM, stale project nudges after 48h, focus alerts below 30, pattern detection after 7 days, and predictive suggestions after 5 days. Persist all data to memory.

Settings reference

All of these are adjustable. Tell your agent to change any setting, or edit state.yaml directly.

SettingDefaultWhat it controls
digest_time"18:00"When the daily digest fires. 24h format.
nudge_stale_hours48Hours of inactivity before a project is flagged as stale.
focus_alert_threshold30Focus score below this triggers a context-switch alert.
focus_alert_window_min20Minutes of scattered behavior before alerting.
pattern_detection_days7Days of history scanned for repeating workflows.
pattern_min_occurrences3How many times a workflow must repeat before suggesting automation.
prediction_min_days5Minimum days of data before predictions activate.
prediction_confidence_min3Minimum occurrences of a routine before it becomes a prediction.
timezonenullYour local timezone. Set once — used for all timestamps and scheduling.