Run Claude Code on your own machine, and drive it from anywhere
Updated 2026-09-02
Claude Code runs on the machine where you installed and signed in to it, which is also where it has to stay. Put a control plane in front of it: a runner on that machine polls Denly for work and pushes status back, so you can queue, approve and review a task from a browser on any device.
Denly is a control plane for coding agents (Claude Code, Codex, Cursor, Grok Build, Antigravity) that run on your own machines and your own subscription: you describe a task, the agent plans it, you approve, it builds locally and opens a pull request. Your laptop or workstation does the work; a browser gives the instructions.
What runs where
| On your machine | On Denly's servers | Outside both |
|---|---|---|
| The agent CLI and its login | The task list, stages and timing | Your Git host receives pushed code |
| The repository checkout | Runner liveness and build version | Your AI provider receives the context its CLI sends |
| Every file the agent reads, edits, builds and tests | Coordination text – encrypted and unreadable when workspace E2E is on | – |
The connection only goes one way
The runner is outbound-only. It polls Denly for work and pushes status back; nothing connects to your machine. There is no inbound port to open, no tunnel to keep alive and no remote exec path, so this works behind a home router or a corporate VPN without a network change.
The cost of that shape: work runs only while your machine is awake and online. You can queue tasks with the lid down, and they start when the runner wakes.
Set it up
- Install Claude Code on the machine that will do the work and sign in to it there.
- Install the Denly CLI:
curl -fsSL https://denly.dev/install.sh | shon macOS or Linux,irm https://denly.dev/install.ps1 | iexon Windows. Both install for your user only and need no admin rights. - Run
denly login. It shows a code, opens your browser, and you confirm the machine once. - Add the project in the dashboard, then run the
denly install <project-id>command it gives you. That clones the repository and starts the runner. - Run
denly doctorto check the platform, the service, the credentials, the runtimes it found and the connectivity.
curl -fsSL https://denly.dev/install.sh | sh
denly login
denly install my-project
denly doctorWhat you can do from a phone
Every decision the agent hands back to you:
- Describe a task and queue it, with an assignee, a runtime and model, attachments, and a task it should wait for.
- Approve the plan, or send it back with notes so the agent plans again.
- Answer the questions the agent asks mid-run.
- Review the pull request and merge it through the recorded gate.
- Open a Live Chat session against the project and watch each tool the agent runs, message by message.
- Read runner health, and restart a runner that has stopped claiming work.
Which agent CLIs this works for
The same arrangement covers Claude Code, OpenAI Codex, Cursor, Grok Build and Antigravity. Availability is per machine: each CLI has to be installed and authenticated where the runner executes, and Denly stores the selection policy rather than your vendor subscription.
Read next
Machines & runners · Live Chat · Run background coding agents on the subscription you already pay for · One dashboard for coding agents across many projects and machines
Denly