Say it or type it
Start an agent with a typed, dictated, or empty prompt. Launch from the project you’re already exploring.
A new way to make things
Big ideas.
Small screen.
Entirely yours.
An AI worker for the things you want to make. Apps, websites, presentations, and whatever comes next. All guided from your phone.
English as a programming language. That’s Englang.
An example brief
Real native UI. Room for your ideas.
“Make a website
for my little bakery.”
Designed for iPhone & iPadPlain English in. Real projects out.Open foundations. Your future.
01 / A little possibility
You have the idea.
Now you have a way.
You shouldn’t need to become a programmer before you can begin making something.
Give your AI a brief. Let it work. Look at what it made. Keep shaping it in a conversation—with the actual project always within reach.
A small-business websiteA book-club appA presentation worth sharingYour next “what if”
02 / A surprisingly capable little workspace
A good conversation
can go a long way.
Simple enough to start with a sentence.
Deep enough to keep going.
01
Start with a sentence.
Tell your AI what you want to make, who it’s for, and what matters. Type it, say it, or attach something that explains it better.
- Typed and dictated briefs
- Photos, documents, and existing files
- Saved prompts you can return to
Tap a screen to take a closer look.
Native app captures · September 9–13
Fictional demo data · UI still carries the Surfcode name
03 / The app, in pictures
The app, up close.
Swipe to explore. Tap to enlarge.
21 native app captures · Demo data · Surfcode UI
01 / 4 native screens
Create
Explore 6 create capabilities
Bring your references
Attach device photos and documents, or reference files already in your workspace.
Dictation, your way
Apple on-device speech by default. Optional OpenAI, OpenRouter, Wispr Flow, or host-helper transcription, with explicit setup and consent.
Review before sending
Edit a transcript before sending. For existing-agent follow-ups, a deliberate mic hold offers release-to-send.
A better vocabulary
Project vocabulary and editable replacement rules help recognition understand the words you use.
Saved prompts & recovery
Reuse locally saved prompts. Keep submission diagnostics and recover text after failed or uncertain delivery; reselect attachments when reusing a prompt.
02 / 4 native screens
Your AI
Explore 7 your ai capabilities
Progress at a glance
Working, ready, idle and attention states, activity labels, timers, and completed-work durations.
A team when you need one
Collapsible subagent trees keep delegated work grouped under its parent, with independent child actions.
Conversations worth reading
Native Markdown, message copying, links, model details and bounded conversation history. Listen to completed answers with system speech.
Keep giving direction
Dictated or typed instructions, reusable quick actions, and a managed follow-up queue keep the work moving.
Try a new direction
Fork a Pi session with its conversation context. Conversations continue independently; project files remain shared.
Know when to come back
Optional named completion notifications open the matching agent preview. Enrollment and notification permission are required.
Manage your workers
Scope by favorite or individual hosts, refresh activity, open the exact terminal, and explicitly manage agent/runtime lifecycles.
03 / 4 native screens
Projects
Explore 6 projects capabilities
A native Projects home
Save repositories, choose an available project, and return to its overview and files.
Start in the right place
Launch an empty, typed, or dictated agent from the project header.
Branches & worktrees
Choose a line of work without losing the surrounding project context.
Recent commits & history
Browse paginated history and open a commit’s changed files. Read another branch without switching its checkout.
Working changes
Inspect staged and unstaged changes. The separate agent Workspace Changes workflow remains a read-only beta, with explicit source-read consent.
Refresh when you need to
Pull to refresh, retain navigation context where supported, and see explicit stale or unavailable states instead of silent fallbacks.
04 / 4 native screens
Code & files
Explore 6 code & files capabilities
Navigate the file tree
Open folders, follow breadcrumbs, and search file paths in a project.
Read actual source
Native syntax highlighting, line numbers, text selection and source copying.
Find your way around
Find loaded text, go to a line, toggle wrapping, or open file and reader details.
Review a change
Added and removed lines, changed-text highlighting and previous/next change-block navigation.
Understand the version
Inspect available working and committed file versions, including parent-specific commit comparisons.
Honest boundaries
Large, binary, unsupported, missing and unavailable files are identified explicitly; a limited read is not presented as a complete one.
05 / 2 native screens
Previews
Explore 5 previews capabilities
Open linked files
Read supported workspace files directly from links in an agent conversation.
Source or preview
Switch Markdown and HTML between their source and rendered presentation.
More than text
View supported images and PDFs alongside plain text documents.
Deliberately isolated HTML
Scripts, external assets and navigation are blocked in HTML previews. This is a safe document view, not a full interactive website test.
Keep your judgment
Review the result before relying on it or sharing it. Generated output still needs a human’s attention.
06 / 3 native screens
Workspace
Explore 6 workspace capabilities
Hosted onboarding
Approved Google accounts can start automatic private-workspace setup, with the agent environment prepared for them.
Connect over the internet
The hosted gateway doesn’t require the user to connect their own Mac or install Tailscale on their phone.
Bring your own host
Save manual SSH hosts, manage device-generated Ed25519 keys, and retain host-key trust checks. Manual SSH does not require a hosted account.
A real terminal
PTY-backed terminal access, tmux sessions, windows and panes, keyboard controls, scrolling and reusable terminal actions.
Control your setup
Explicit integration installation and runtime updates, host preferences, and guarded hosted-workspace deletion.
Clear privacy choices
Device Keychain for supported credentials, local-first speech, deliberate cloud-transcription consent, and optional push enrollment. Projects and conversations are not automatically made public.
04 / The part that really matters
Make something.
Don’t give it away.
Not to a platform.
Not to a single AI.
Not to a format you can’t leave.
Your Git hostYour next AIYour memoryYour domain
The constantYour
project.
project.
Your choices can change.
The work comes with you.
Own the whole project.
Keep the source and files in GitHub or another Git host you choose. Not just a preview you can’t take with you.
Your website. Your rules.
Build self-sovereign websites. Keep the source, use a domain you control, and choose where your website runs.
Choose your next AI.
A capable successor can pick up your files and saved context. Your project’s future needn’t depend on one model or vendor.
Open foundations. Open options.
Open-source systems, standard files and Git help you avoid lock-in—and keep moving as the tools around you change.
Keep what you learn.
Ask your agent to maintain preferences, lessons and decisions as you iterate. Own its memory and your conversations, so the next round can build on the last.
Ownership is a way of working, not a magic export button. Hosting, Git accounts and agent handoffs may need setup. Retain and back up memory and conversations; don’t publish private history or secrets by accident.
05 / Less explaining. More showing.
A little tour.
A bigger possibility.
Fourteen moments inside Englang.
Two minutes and forty-four seconds.
06 / A few good questionsBefore you
Before you
start dreaming.
Can I try Englang today?
Englang is currently in a private pilot, and hosted workspaces require an approved account. If someone shared this with you, ask them about joining. Public signup isn’t available yet.
Do I need a Mac—or know how to code?
The hosted path doesn’t require connecting your own computer. You can begin by describing what you want in plain language. You’ll still need to review the work, and complex projects may need technical help. Advanced users can connect an existing SSH host instead.
Can it really make more than apps?
Use your AI worker for websites, apps, presentations, documents and other file-based projects. The examples on this page are starting points, not guarantees of a finished result. Like any AI-created work, the output needs your review; publishing or distributing it may take additional setup.
What does “own the memory” mean?
Keep the notes, preferences and decisions your agent uses, alongside the conversations you choose to retain. Ask it to maintain that context as the project evolves. It can work better with those lessons without retraining its underlying model. You decide what to keep, and should back up that history—not assume every conversation is already saved in Git.
Can I change hosts or AI agents later?
Real source files, Git history and saved context make that possible. A successor agent needs the right context, tools, permissions and setup; migration isn’t necessarily one click. Open foundations preserve choices, not a guarantee that every future tool will work unchanged.
Why do the screenshots say Surfcode?
Englang is the proposed name, from “English as a programming language.” The screenshots are real native views from the current Surfcode app, captured September 9–13 with fictional data. They are not mockups of a shipped rename. Some earlier controls and layouts have since evolved.
Your ideas. Your direction. Your future.
The next thing you make
could start right here.
One phone. A few words. Something that’s yours.
See what’s possible