For developers and small teams using several CLIs
Nothing ships without your review.
Claude Code, Codex, Copilot and Antigravity on your own machines and plans. One CLI reviews another, you approve the push.
- Free for 1 agent
- 14 days of Pro free
- No credit card
- Set up in 5 min
One CLI writes, another reviews. You decide what ships.
SideHub’s chain of control, from the task to the push: every step has an owner and shows what it cost.
Demo: Claude Code implements a task on a VPS, Codex reviews it, the work is approved from a phone and pushed, then the cost of each CLI shows in the cockpit
One CLI codes
Claude Code, Codex, Copilot or Antigravity: the one you pick implements the task on your machine, in the real CLI, with your plan.
Another reviews
Claude Code or Codex reads the diff in read-only mode, by default the one that did not write it, and ranks its remarks. You keep the ones that matter.
You approve
The “To review” card reaches your phone: diff, remarks, Approve or Request a fix.
SideHub pushes
Only the HEAD you approved goes to the remote branch, never forced.
Cost of the two tasks in the video, estimated API equivalent: priced at the models’ API rates, from the tokens read in the CLI transcripts
One CLI per app. No guardrail before it ships.
Each vendor drives its own CLI, and only its own. Between the agent and your branch there is only you, when you remember to look.
Each CLI in its own app
Claude in the Claude app, Codex in ChatGPT, Copilot in GitHub Mobile: three histories, three ways to approve, no overall view.
Nothing between the agent and the branch
The agent commits and pushes right away. Nobody read the diff, and no other model checked its work.
What a task costs stays fuzzy
Every plan has its limits, but none tells you what a task cost or which CLI does best on your code.
Agents run on your side. SideHub drives them.
The cockpit decides what to run, where and when. Your machine runs it, and keeps the code, the secrets and the model traffic, under your own licenses.
Control plane: The cockpit, on desktop and phone
- Orchestrate
- Schedule
- Approve
- Measure
Goes through here
- Orders, statuses and end reasons
- Token counts
- Terminal output, relayed, not stored
Execution plane: SideHub Agent
Installed on your machines: laptop, VPS, VM or build machine, wherever your code lives
Stays here
- Source code
- Secrets and .env
- Model calls
- CLI sessions
Linux, macOS and Windows, outbound connection only
A new Claude Code session, started from your phone.
Not a remote view of a conversation left open at your desk: pick a machine, launch the CLI and give it the task. It asks when it needs you, and the result lands in Memory.
SideHub or Claude Code Remote Control?
- Your machine, your CLI, your login
- Answer from the cockpit when it waits for you
- Results kept in Memory for the next session
By day, you drive them. By night, they work on their own.
An automation runs one prompt or a multi-step workflow, each step with its own CLI and model. Scheduled, it starts on time on the machine you pick, and anything that needs your approval waits until you wake up.
6 workflows, 5 of them scheduled, run every day
SideHub’s blog and its maker’s projects are produced by its own agents.
Read the code of what runs on your side
The agent is MIT licensed. Check what it sends, build it yourself. Agent on GitHub
Five minutes to your first run
Install the agent on your machine
A standalone binary. It opens an outbound connection to the cockpit and shows up online.
Give it work
A session from your phone, a multi-step workflow or a scheduled automation that runs overnight. Each step picks its CLI and model.
Approve what matters
Before a sensitive action, the run stops and asks you. You answer from the cockpit or Telegram, and it picks up where it left off.
Follow every run
Real status, end reason, live terminal, tokens and cost per model.
Free to start, €10 when your agents work overnight
Your Claude, ChatGPT, Google AI or Copilot subscriptions stay yours. SideHub only bills the cockpit.
Starter
0 € forever
Try SideHub with one agent on one machine.
- 1 workspace
- 1 agent
- Remote terminal and tasks
- Workflows launched by hand, without secrets or MCP servers
- 50 Memory documents, 100 MB
Pro
10 € / month
Solopreneurs running several projects and agents.
- Unlimited workspaces
- Up to 5 agents
- Automated workflows: schedulers, triggers, webhooks
- Secrets and MCP servers for workflows
- Shared Memory, 10 GB
- Semantic search (optional, with your OpenAI key)
Business
15 € / member / month
Teams sharing agents, automations and memory.
- Everything in Pro
- Billed per member, Viewers free
- Unlimited agents
- Telegram bot
- Unlimited storage
Enterprise
Companies that want a pilot on their own infrastructure.
- 30-minute demo
- Pilot deployed on your side
- Your company’s Copilot licenses
- SSO (coming soon)
Without a subscription after the 14 days, the account moves back to Starter. Nothing is deleted.
Frequently asked questions
Do I have to code with Claude Code and review with Codex?
No, that is the video’s example. Each task runs on the CLI you pick: Claude Code, Codex, Copilot or Antigravity. Reviews run on Claude Code or Codex, the two CLIs SideHub can keep read-only, by default the one that did not write the code: Codex can write and Claude Code review, or the other way round.
Does my code go through SideHub?
No. It stays on your machine. SideHub receives orders, run statuses, token counts and terminal output, relayed live to the cockpit and not stored (except the last lines of a failed run).
Do I need API keys?
Not for the agents: they use the CLIs as you configured them, with your subscriptions or your keys, and SideHub is not a proxy to the models. Only two cockpit options, the assistant and Memory semantic search, use your own key if you turn them on.
Can we use our company’s GitHub Copilot license?
Yes. The agent runs GitHub Copilot CLI with the account signed in on the machine: sign it in with your organization’s account, and its data and model policies apply. SideHub shows the account, plan and organization of each machine, and flags the ones on a personal subscription or an account outside your organization. Traffic goes from your machine straight to GitHub: SideHub never sees it.
Do I need to open a port or a VPN?
No. The agent opens an outbound WebSocket connection on port 443. No inbound port is needed.
What happens after the 14 days?
Without a subscription, the account moves back to Starter (1 agent, free). You can change plans at any time from Settings.
Can I read the agent’s code?
Yes. It is open source under the MIT license, at github.com/vincentaugerot/side_hub_agent. You can build it yourself.
Your first agent online in five minutes
Create your account, paste the command on your machine, hand over a first task and review it before it ships.