To sieve: the art of separating and discriminating

43 thingsarrived.Two matter.

OnMe turns screenshots, links, and scheduled checks into PURSUE, WATCH, or SKIP. Personal, market, and project context stay in their lanes.

See today’s decisions ↓

20 founding accounts. Bring ChatGPT, Claude, Cursor, or another execution host that passes OnMe’s heartbeat test.

$10 billed monthly until canceled. By subscribing, you agree to the Terms and acknowledge the Privacy Notice.

Today’s decisions43 evaluated · 2 delivered

SHIPConfirmed: OnMeDependency match

React Native security release affects the build.

Pursue

PROTECTTomorrow · 9:00

Leave by 8:20 for the appointment.

Interrupt

22 silent19 skipped

Everything else stayed out of your day.

Trace

01 / The loop

Information has to earn the interruption.

Your existing AI gathers evidence. OnMe owns the rules around it: where the result belongs, what it may affect, how much reasoning it earns, and whether you should see it now.

Define the permission

Pick a trigger, jar, allowed effect, interruption policy, budget, expiration, and destination.

Let your agent gather

ChatGPT, Claude, Cursor, or another capable host collects the evidence and sends it to OnMe.

Get the decision

OnMe returns PURSUE, WATCH, or SKIP, plus the reason, evidence, cost class, and next scheduled time.

Correct the memory

Accept it, change it, reverse it, or add the eventual outcome. Cleared work leaves active attention.

02 / The wedge

Your AI can find everything. It should not interrupt you about everything.

OnMe is not a second chatbot and not a prettier queue. It is the decision boundary between what your agents can discover and what earns the right to change your day.

Generic scheduled AI

Produces another briefing

Mixes personal and project context

Treats every result as an alert

Forgets whether its advice helped

OnMe

Commits to pursue, watch, or skip

Enforces jar and project permission

Can finish without interrupting

Learns from acceptance and reversal

03 / Founding offer

Choose who executes. Keep the same permission boundary.

Use an agent you already pay for, or let OnMe run the bounded background loop. Both plans keep decisions tied to explicit Tracks.

OnMe / Sieve

Customer-hosted AI through your certified Codex, Claude, Cursor, or another compatible host. OnMe makes zero calls with OnMe provider credentials.

10 ACTIVE TRACKS / TASKS100 EVALUATED RUNS / MONTHDETERMINISTIC + ON-DEVICECUSTOMER-HOST EXECUTIONDECISION HISTORY + FEEDBACK

$10PER MONTH
CANCEL ANYTIME

Your compatible host and its subscription remain yours.

Renews monthly until canceled. Terms · Privacy

OnMe / Hosted

Reliable OnMe background execution when your desktop host is unavailable, within a hard monthly provider-spend ceiling.

100 HOSTED DECISIONS10 DEEP ESCALATIONS20 BROADER-WEB CYCLESMAX 3 SEARCHES / CYCLENO SILENT OVERAGES

$20PER MONTH
CANCEL ANYTIME

At the spend ceiling, OnMe uses your connected host or returns needs_executor.

Renews monthly until canceled. Terms · Privacy

Is this another AI assistant?

No. Your chosen agent does the tool work. OnMe owns the task contract, permission boundary, verdict, interruption directive, and decision memory.

Will every ChatGPT, Claude, or Cursor plan work?

No. Provider plan names are not treated as capabilities. OnMe records a host as compatible only after it completes a real authenticated heartbeat.

What happens to completed work?

It leaves active attention. OnMe retains a compact trace of the recommendation, accepted or reversed verdict, jar, reason, timestamps, cost, and eventual outcome.