Without Ondago Run paused
3 hrs
of nothing happening
You close the lid mid-run. The session waits exactly where you left it, and tomorrow starts by rebuilding the context you already had.
Continuous agentic development
Your laptop can stop.
Your agentic work
doesn’t have to.
Ondago hands your active coding-agent session, with its context, instructions, memory and progress, off to a secure cloud machine. The agent keeps working while you commute, meet or close the lid. You steer it from your phone and pull the session back when you return.
Works with Claude Code, OpenCode and Codex · private secure cloud · no restart, no re-reading the problem.
One session, four steps
The problem
Your agents are autonomous. Your setup isn’t.
A coding agent can work for an hour unattended. But it runs on your machine, so the moment you leave, the work leaves with you. Every outcome below costs the same thing: time you paid for and did not get.
1
You wait at your deskTwenty minutes into a run you have somewhere to be, so you sit and watch a process that does not need you.
2
You lose the runClose the lid and the agent stops mid-task. You come back to a half-finished branch and start the context over.
3
It sits blockedThe agent hits a question after ten minutes, then idles for the next fifty because nobody was there to answer it.
And the runs keep getting longer
Agent sessions are getting longer.
Coding agents have moved past one-shot prompts. They plan, build, run the tests, read the results, fix what failed, and loop again. Each loop produces better work, and the full run takes hours.
01Plan the changeThink
02Write and update codeBuild
03Run tests and inspect resultsCheck
04Fix what failedImprove
↻ Repeat until the work is ready
The solution
Turn waiting time into finished work.
The agent is not faster. It simply gets uninterrupted time to finish the work you already asked it for.
With Ondago Run complete
41m
of work you did not wait for
The same run continues in the cloud, reaches you once for an approval, and is finished and merged by the time you sit back down.
Keep the run moving
Meetings, commutes and closed laptops stop being reasons for a stopped run.
Know when to step in
See the moment an agent needs a decision, and answer from the device already in your hand.
Move the whole session
Context, instructions, memory and the working tree travel together, not just the diff the agent produced.
Keep momentum, not just compute.
Keep the agent moving through commutes, meetings, and breaks, so progress continues while your attention is elsewhere.
Turn commutes and downtime into productive agent time. Features keep moving while you step away.
Share the live working session, not a long recap. Teammates arrive with the complete working state.
If you are sick or unavailable, hand the live session to a teammate. They continue with the same agent, files, and context instead of starting over.
How it works
Local when you want it. Cloud when you need it.
-
Step 01
Work as you already do
Run your agent locally, in the tools you already use. There is no remote environment to prepare first.
-
Step 02
Hand the session off
Run
ondago lift. One command moves the live session, with its context window, instructions, memory and working tree, into a private cloud environment. -
Step 03
Stay in the loop from anywhere
Watch progress, answer the agent’s questions, approve sensitive actions and queue new instructions from your phone.
-
Step 04
Bring it back
Run
ondago pull. The session and everything it accomplished return to your laptop, and the agent resumes on the exact step it was on.
Handoff, in practice
$ ondago lift
› capturing session state…
› context 128k · memory · 12 changed files
› session live in cloud · streaming to mobile
$ ondago pull
✓ session restored · 41m of work merged locally
The agent never restarts and never re-reads the problem. It resumes on the exact step it was on.
From your pocket
Command the run from anywhere.
A blocked agent becomes a notification instead of an hour of dead time. Approve, redirect or extend the run in a few taps, then read the summary when it finishes.
- Live progress: files touched, tests run, the step it is on right now.
- Unblock the agent the moment it asks, instead of an hour after it asked.
- Approve sensitive actions explicitly, never by default.
- Queue the next prompt before you are back at your desk.
Security posture
Your code stays yours.
The session is sealed on your machine before it travels. What the service holds is ciphertext, which is a property of the design rather than a promise about our behaviour.
Sealed on your machine
Your session is encrypted before it leaves. We store ciphertext and hold no key that opens it, so there is no server-side search, preview or summary. Session lists decrypt in your own client.
Isolated environments
Each run gets a private environment of its own, destroyed when the session ends. It never holds the key to your account.
Scoped credentials
Provider keys are injected for one run and never persisted. Sensitive actions wait for your explicit approval instead of assuming it.
In plain English
Questions you may already have.
Do I have to move all my work to the cloud?
No. Ondago moves one running session at a time. Keep your normal local workflow and hand off only the run that needs to continue without your laptop.
Do I need to restart the task?
No — that is the point. The conversation, working tree and uncommitted changes move together, so the agent never restarts and never re-reads the problem.
What happens when the agent needs me?
You get a notification on your phone with the exact question or action awaiting approval. You answer in a few taps and the run continues immediately.
Can I bring the session back to my laptop?
Yes. Run ondago pull to bring back the full session and the latest changes, then continue in your normal local setup without losing context.
Where does the session actually run?
In a private cloud environment created for that run and destroyed when it ends. It is sealed to keys your machine holds, and it never holds the key to your account. See security posture.
Which agents does it work with?
Claude Code, Codex and OpenCode to start, with more coming as part of early access.
Is this only for technical teams?
The people running the agent are, but the outcome is not: less paid time spent waiting, more runs finished, and visibility into work in progress for everyone who depends on it.
Early access
Stop trading your time for agent uptime.
We are inviting a small group of individuals, builders and teams that run long agent sessions and want fewer interruptions in the day.
Request early access