Are you a software engineer,
or a frontier engineer?

You can already spin up ten agents at once. The hard part is controlling them: different UIs depending on model or provider, seeing what each one is doing and steering the ones that drift, sitting on prompts waiting for others to finish, using the CLI one minute, a chat app the next, scheduling that is hit and miss…

Frontier removes all of this friction so you can focus on what actually matters, advancing on all fronts.

Download See how it works →

Frontier’s desktop app runs on macOS, Windows & Linux — open this page on your computer to download.

macOS Windows Linux
FrontierAlphaSpaces
Check-inTranscriptScheduledCommitsSettingsAwaiting input·11m 00s
Summary 11m 00s ago

Assessment

The split into ingest / api / worker is sound. One sharp edge: the worker and the api both write sessions, so the boundary has to own a single writer — or you reintroduce the exact race we just fixed in the token store.

Trade-offs

  • A dedicated session service is cleanest, but adds a hop.
  • Folding writes into the api keeps the topology smaller.
Direction

Answer below, or type a direction in plain language.

Questions 11m 00s ago
1## Proposals
2
3### Who owns session writes after the split?
4
8
9Recommendation: api owns the write, the worker calls it.
10One writer, no new service, the boundary stays honest.
Frontier running several spaces in parallel: a live check-in with summary, questions, and a nested subagent tree.
The manifesto

What is frontier engineering?

AI has drastically changed how we work.

You’re still shipping features, but you’re also planning the launch, clearing action items automatically captured from your meetings, reviewing the reports that were autonomously written overnight and nudging an Unreal Engine side-project forward, all while planning a multi-leg international holiday. Frontier engineers advance on many fronts at once — a step-change in productivity across everything you do, not just the code.

You’re a new breed of exploratory creator, a skilful delegator, your producing value faster than anyone ever has, and your limited only by the scope of your ideas and the quality of your tools.

Frontier is the operational infrastructure for frontier engineers.

The problem

You can run ten agents. Controlling them is the hard part.

Waiting was the old problem, and you already beat it by running agents in parallel. The new problem is what parallel actually feels like: terminals everywhere, subagents you can't see inside, half of them idle waiting on you, and a finished prompt you're holding because it isn't safe to send yet. More agents, more chaos, unless something owns the coordination.

Five terminals, no control

You're the router and the bottleneck

  • agent-1 finished 4m ago, you didn't notice
  • agent-2 drifting, no way to see inside
  • prompt ready, held in the CLI, not safe to send
  • agent-4 idle, waiting on you
  • You're the monitor, the router, and the thing everything waits on.
With Frontier

Every front visible, each on the right model

  • auth-refactor running · machine-02
  • data-pipeline queued → runs when a slot frees
  • pricing-page needs a decision → you answer
  • launch-imagery running · image model
  • You answer the one that needs you. Frontier routes the rest.
See how Frontier handles it
Why it matters now

Build and Adapt Faster

The way you work and the tools you use will keep changing at a dramatic pace over the next few years. Frontier is built to be the one constant through it all. Not by being the best tool at any given moment, but by being the one that lets you follow the industry the fastest.

New interaction patterns emerge every week. Some will last, most won’t. Frontier lets you fold the good ones into your workflow in a few iterations with a model: you take the best and leave the rest. Your advantage as an engineer is now how fast you can pull the best ideas in the industry into the way you work.

You don’t read the docs. Your agent does.

Reshaping it is a conversation. You say what you want, your agent reads the live API over MCP, writes the code, and hot-reloads it. An extension is small enough that this just works.

  1. 01

    Describe it

    “A Standup extension that pulls today’s commits, with its own view.”

  2. 02

    It reads the live API

    Real types and working examples, handed to the model over MCP. Not a docs site you study.

  3. 03

    Writes & hot-reloads

    The extension compiles and the server and UI reload in place. No restart, no boilerplate.

  4. 04

    You watch it run

    Tweak it by talking. The extension improves itself while you’re using it.

No bait-and-switch

Free today, free forever.

Every extension on this page is free to use today. The core extensions we ship (Spaces, Meetings, Assistant, Chat, Files, and more) will keep evolving and stay free to use. Always. Forever. That’s a promise.

Always free
  • Every core extension we ship
  • The host and the scheduler it all runs on
  • Running it on a single machine or an entire fleet
How we’ll make money, eventually
  • Optional hosted instances (we run it, you skip the setup)
  • A marketplace of premium or third-party extensions
  • …and the core stays free regardless

We’re telling you the plan now, before you’ve put anything in, so it can never become a bait-and-switch later.

Why Frontier

Why not just Claude Code, Cursor, MyClaw or opencode? Why not all of them?

You probably already use one of these, and that’s fine. Claude Code works from the terminal, Cursor is a paid AI editor, and opencode is the open, terminal-first take. They’re all good at writing code. Frontier is the free, self-hosted layer below them, and it isn’t only for code. Here’s where each lands.

And you don’t have to choose. Frontier lets you build your own workflows from the best ideas across all of these tools — and whatever ships next — then dictate exactly how you use it. Bring the models from any of them, too: nothing here locks you in.

The surface A terminal agent An AI code editor An open terminal agent A workspace you rearrange, on any machine
Models Claude only A few to choose from Any model The right one per front and per subagent
Subagents Invisible. You can’t intervene. Agents, but inside the editor Visible sub-sessions you follow Nested spaces you watch and steer
Shaping it Hooks and MCP, around their core Their editor, their AI Open source. Fork it all. Every feature is a swappable extension
Beyond code Code, in the terminal Code, in the editor Code, in the terminal Meetings, an assistant, dictation, and extensions you build
Price Free tool, on your Claude plan or API usage A paid subscription, on top of model costs Free and open source Free and self-hosted — bring the agent you already pay for
FAQ

Questions, before you ask.

Is this another editor like Cursor or VS Code?

No. Frontier sits a layer below an editor. It's the infrastructure that runs sessions across machines, persists state, and routes messages. An editor is one of the things you could build on it. The Spaces workflow you see here is built on Frontier, not baked into it.

Do I have to use the Spaces workflow?

Not at all. Spaces is the reference extension, the one that grew out of the daily grind of waiting on a single chat. You can retune it, fork it, or write your own workflow against a small, design-focused API.

Which models does it support?

Claude, GPT, Gemini, and local models via Ollama. You choose per workspace and can switch at any time without changing extension code, so different fronts can run on different providers at once.

What happens if I send a prompt while an agent's still working?

It queues. Frontier decouples submitting a direction from running it: your prompt waits for a free workspace and dispatches the moment one frees, latest one wins. You write what you want and walk away; you never sit holding a prompt for a safe moment to hit enter.

Can I use it yet?

Yes — download the app for macOS, Windows, or Linux and double-click; it works the moment it opens. From inside it you connect a host and add machines over SSH. It's free and yours, early and evolving fast, so expect it to get better under you.

Free to run

Stop waiting.
Advance on every front.

Download the app, double-click, and you’re in. From there, connect a host and add machines over SSH — no terminal, no setup.

Download

Frontier’s desktop app runs on macOS, Windows & Linux — open this page on your computer to download.

macOS Windows Linux

Free and yours. Your code never leaves your machines.