Workstation for your agents.A window for you.
Your agents run on hardware you own and keep working after you close the browser. Reopen from any device to steer the fleet — and read every change before it lands.
- Free and open source
- Bring your own agent
- Runs on your hardware
Run a fleet. Stay in control.
Ten agents at once, wired into your systems, woken by events, and reviewed change by change before anything lands.
Run ten agents at once. The board sorts the fleet into who's blocked, who's running, who's done — and surfaces the one that needs you.
Empower
Wire an agent into GitHub, Postgres, Sentry, Stripe, Discord or any MCP server — it sees and acts, and the credential stays in your sandbox.
- a push
- a Sentry alert
- a Stripe payment
- inbound email
- a chat message
- plain cron
a guard you write decides, then a fresh agent session runs and lands like any other.
Automate
Agents that start themselves — on an event you choose, under a guard command you write, each run a fresh session you can watch.
Supervise
It plans first and you approve; finished work waits on its own branch until you have read every hunk of the diff.
Delegate
Give the workspace a server of its own and hand off end-to-end operation — the sandbox runs on hardware you own, the platform off the command path.
Your code never leaves your machine.
An agent with your keys is only safe if you own where it runs. Here is exactly what sits on each side.
Your machine
Where the sandbox runs.- Your repositories and working tree
- API keys, .env files, database passwords
- The agents, their sandboxes, their history
- Every file an agent reads or writes
The intentic platform
Everything it stores, in full.- Your email address
- Your sandbox's URL
No code, no keys, and no way to command your agents.
Your browser holds the token that drives the sandbox; the platform never does. All of intentic is MIT on GitHub, platform included, so you can check.
How ownership works in the product →A whole fleet, on the subscriptions you already pay for.
Ten agents sounds expensive. It isn't: each runs on a plan you already have, on hardware you already own.
- subscription
Claude
Opus, Sonnet and Haiku, on your Claude plan - subscription
Codex
on your ChatGPT plan - subscription
Grok
on your SuperGrok plan - subscription
Kimi Code
on your Kimi Membership - subscription
Google
Gemini, Claude and GPT-OSS, free on a Google sign-in
- No per-token metering. No markup on your model usage.
- No rented cloud compute. Agents run where you run them.
- Free, whole: every sandbox, capability and shared workspace included.
Artur Kurowski
Full-stack engineer, DevOps, and 15+ years of shipping production systems.
I built intentic because I wanted a real workspace for my own agents. Not a chat window, and not somebody else's cloud holding my source. It runs on my hardware, and now on yours.
More about who builds this →Why trust intentic?
Built by a verifiable person, and built so that trusting the person is optional. The platform holds your identity and a URL, with no path to your code, your keys, or your sandbox.
What the platform actually holds →Open source first
All of intentic is MIT on GitHub — the parts that touch your credentials and the platform behind them. No hidden binaries, no telemetry. Read what executes on your hardware before you run it.
Read the source →It builds itself
Agents running in this product wrote most of it, in public, one reviewable commit at a time. The fleet on this page is what shipped the page.
Read the commit log →Honest about its age
No case studies, no testimonials, no logo wall. The project is young, and what you get is a real sandbox rather than a trial. Read what it does before pointing it at something you care about.
One command, and an agent has a home.
Sign in, then paste one command on the machine that should host it. The workspace opens the moment it reports in.
Sign in with Google
No forms, no card. It stores your identity and the sandbox's URL, nothing else.
The sandbox is waiting
Made and named for you, with a private tunnel under intentic's own domain. No Cloudflare account needed.
Paste one command
A one-liner starts the sandbox on your machine. Docker installs if missing, with your say-so.
$ curl -fsSL https://intentic.dev/connect | shNothing deployed, nothing exposed. Just a workspace your agents can call home.
Questions, answered plainly.
Including the honest ones: what's free, what's stored where, what isn't ready yet.
What it is
The product, how it differs, and what it runs alongside.
What happens when I close the browser?
Nothing, to the runs. The agents live on your machine, not in the tab — terminals survive the disconnect, and turns finish without you.
Reopen from any device, including a phone, and the same fleet is there, sorted by who now needs you.
How is this different from a custom GPT or a .md instructions file?
Those are a prompt on a generic assistant: nothing installed, no reach into your code, a blank context every time.
An agent here gets a container with its tools really installed, credentials for your repos and databases, and its skills loaded every run.
How does this compare to Conductor, Cursor, OpenCode or Nimbalyst?
Mostly it doesn't compete. Claude Code, Codex and OpenCode are agent harnesses, and intentic runs all of them.
Local orchestrators like Conductor share the ownership stance. Against a cloud platform, the only real question is whose computer holds your source.
Every comparison, with the case for the other product →Can I run a separate agent for each job?
Yes, that's the intended shape: a migrations agent, a release captain, a support triager, each with the environment and access its job needs.
Automations wake them on a schedule or an event, each run a fresh session with its own transcript.
Specialize a sandbox until it does a job alone →Should I start a new sandbox or add a new persona?
Go by what's different. A sandbox is a whole workspace of its own — its tools, the systems it's connected to, its budget and its code. Start a new one when the job is genuinely different, or when you want that work kept apart with its spend on its own tab.
A persona is who a sandbox is when it posts to the outside world: it ties together the logins that are the same someone, so the agent speaks as that account and no others. Add one — no new sandbox needed — when the work is the same but you want a different face for it, like posting as a brand account alongside your own.
Specialize a sandbox, then grow a team →Which AI models does it use?
Claude Code (Opus, Sonnet, Haiku), Codex, Grok, Kimi Code or Google, picked per conversation. Your provider, your account, your usage.
Getting started
What you need, what it costs, what it can touch.
What do I need to run a sandbox?
A machine with Docker and a Google account. No open inbound ports, nothing deployed. Docker installs itself if missing, with your confirmation.
Do I need a Cloudflare account?
No. The tunnel is provisioned under intentic's own domain by default. Bring your own zone if you prefer; the token is used once and never stored.
Is any of it paid?
No. Every sandbox, every capability and team sharing are free — no tiers, no limits, no card.
All of intentic is MIT on GitHub, platform included. You pay your own model provider, directly.
Can the agent break my stuff?
It starts in plan mode: it proposes, you approve. Every change is a diff you can discard, and Dockerfile changes need explicit approval.
Stricter and looser permission modes are one click away.
Is it production-ready?
The app is new and says so, but it isn't a demo: what you get is a real, full sandbox.
Every agent starts in plan mode and every change lands as a reviewable diff, so you build trust with the wheel in your hands.
Ownership & privacy
Where your code, secrets and data actually sit.
Where does my code live?
On your machine. The sandbox runs where you start it, and your browser reaches it over a private tunnel.
The platform stores your identity and the sandbox's URL. It never relays your files.
Can intentic read my secrets?
No. Your capability credentials live inside your sandbox and the platform never receives them. The connection secrets it does hold — the ones that pair a browser to a sandbox — are encrypted at rest with AES-256-GCM.
Secret files like .env are denylisted from the file relay, so they never leave the sandbox.
Is it open source?
Yes, all of it. intentic is MIT on GitHub — the sandbox, the CLI, the workspace and the hosted platform, developed in one public repo — so you can read every line that touches your code before you run it.
The platform adds accounts and the connection layer. It stays off the command path, and now you can check that for yourself rather than take our word for it.
Can I export or delete my data?
Settings → Export downloads everything the platform stores about you as JSON. Account deletion cascades sandboxes, sessions and grants.
Still unanswered? Issues are public and get a real reply.
Open an issue →Put ten agents to work. Come back whenever.
One command to a live sandbox on your hardware. Free — and nothing lands until you have read the diff.