Fredrin01
This desktop
Free on every plan
The agent runs inside the app, in a real terminal you can watch and type into. Nothing to pair.
macOS, Windows & Linux
Cloud agents,
on machines you already run
Pair a Mac mini, a Linux box or a VM once, then send it tickets from the board. The work keeps moving with your laptop shut.
The fleet
One board. Every machine you own.
A Mac mini under the desk, a Linux box in a rack, a teammate's shared machine. Each ticket picks where it runs, and the card keeps its live terminal wherever that is.
A Fredrin board beside its list of remote machines. Tickets are dispatched from Backlog to a paired Mac mini, a Linux server and a teammate's shared machine, run there, and land in Review for a human to look at.
Remote agents, shipped
Remote AI agents, paired in one command
One daemon on the machine dials out to Fredrin, so there are no inbound ports and no VPN. It serves every workspace you pair it to, and each ticket it takes runs in its own git worktree with the live terminal streamed back to the card.
Remote agents are included from the Pro plan up. Running Workers on this desktop is free on every plan. See what each plan includes.
Three places a ticket can run
The choice is made per ticket in the Runs on picker, not per account, so a copy fix can run here while a migration runs on the box under the desk.
02
A paired machine
Pro and above
Yours, or one a teammate shared with the workspace. Sessions park between turns and keep running with your laptop shut.
03
Fredrin Cloud
$5/mo included
An instance per ticket that boots when you press Build, with nothing to pair. Every plan includes $5/mo of it, and paid plans meter by the second past that.
By construction
Two things about a remote run that are true because of how it is built, not because of a policy.
- command to pair a machine
- 1command to pair a machineThe installer is one line. Re-running it upgrades the agent and keeps every pairing.
- inbound ports
- 0inbound portsThe machine dials out. No VPN, no tunnel, no firewall rule.
What a remote run actually does
A remote session is always on. It parks between turns instead of dying, so you can answer a question from a phone an hour later and pick up the same conversation.
01
You press Build
Set the ticket's Runs on to a machine and build. The daemon on that machine claims the job on its own credential; nothing is pushed to it over SSH.
02
It cuts the branch in its own worktree
The default branch is fetched fresh, the ticket's branch is cut from that, and it is checked out in a git worktree of its own. Ten tickets on one box never touch each other's files.
03
The real CLI starts, in a real terminal
Claude Code or Codex runs in that directory in a real PTY. The live terminal streams back to the card, and you can type into it from the board.
04
The turn ends and the session parks
A finished turn moves the card to Review and parks the session. A question parks it in Blocked and rings for you. Nothing is killed for being idle.
05
You reply on your own schedule
Answer from the desktop app or the board in a browser, hours later if you like. The same session picks the conversation up where it stopped.
06
Completing the ticket closes it
Merging or completing the ticket closes its remote session within seconds. One that is simply forgotten is closed by a three-day age cap.
What leaves the machine, and what never does
Fredrin is a board around your own repo, keys and agents, not a cloud that holds your source. The whole boundary fits in three rows and one drawing.
Stays on the machine that runs the ticket
Always. There is no setting to find.
- Your source code: a local git worktree on that machine
- Your AI subscription and keys: agent traffic goes to your provider, never through us
- Any GitHub secret of its own: Fredrin sends the dispatcher's credential with each job, so the box keeps none
Syncs to run the board
Metadata only. Never code.
- Tickets, columns, comments: the kanban itself
- Run status: exit codes and token counts
- Branch names, PR links, CI status: the diff stays on your git host
Travels through Fredrin when a ticket runs remotely
Dispatching to a machine is the consent.
- The live terminal and its transcript: so you can watch a machine you are not sitting at
- Secrets scrubbed before anything is written: and readable by your project
- Delete stored history: erases it, any time
Not on the price list yet: single sign-on and SCIM, role-based merge governance, a tamper-evident audit log and a SOC 2 report are on the roadmap. If one of those is the blocker, tell us which. The full boundary is on the security page.
Hosted compute
Cloud agents, hosted by Fredrin
Every plan includes $5/mo.
Set a ticket's Runs on to Fredrin Cloud and it gets an ephemeral machine of its own: it boots when you press Build, clones the branch, runs the same agent, parks when idle and sleeps while it waits on you. Nothing to install and nothing to pair. It needs a GitHub repository connected to the project, because that is where the branch is pushed.
Every plan includes $5/mo of hosted compute. Past that, paid plans meter by the second and the free plan stops at the allowance. Your own machines stay unmetered, and they are still the default.
Requestingan instance is asked forBootingit clones the branchReadyconnected and runningAsleepwaiting on you, meter stoppedDestroyedthe next build starts fresh
Questions & answers
- Do I need to leave my laptop on?
- Not for a ticket running on a paired machine. The session lives on that machine and parks between turns, so your laptop can be shut and the card keeps moving on every device you open the board from. A ticket set to run on this desktop does need this desktop.
- What does it cost?
- Remote agents are included from the Pro plan up. Running Workers on your own desktop is free on every plan. In both cases the agent runs on the coding subscription you already pay for.
- How is Fredrin Cloud billed?
- Every plan includes $5/mo of hosted compute, shown as a live balance in the app. Past that, paid plans meter by the second against the same balance; the free plan stops when it is spent and resets with the billing month. Agents on your own machines are never metered.
- Can my team share a machine?
- Yes. A paired machine can be shared with the workspace, and then it takes every member's tickets. Sharing a machine is handing out code execution on it, and the switch that turns it on says so.
- Which coding agents run remotely?
- Claude Code and Codex, the same two the desktop runs, with more on the way. The CLI has to be installed and signed in on the machine, the same as it would be on your laptop.
- Is this the same as running the agent over SSH?
- No SSH is involved. The machine dials out to Fredrin, so you open no port and no tunnel, and the terminal you watch on the card is the live one on that machine. The same setup works on a home server, a spare Mac or a VM you rent.
Setting a machine up end to end, including Windows and the GitHub account a machine's jobs run as, is in the pairing guide.
Put a machine to work
Free to start on this desktop. Pair the first machine when you are ready.