Early beta · Download and install directly from your terminal

GOOD WORK, MADE TOGETHER

Tell Miles what you want to build. Your AI team brings you a clear plan and work you can review.

On your computer or server · Your data, under your control.
Self-hosted workspaceYour AI accountsNo FellowBy cloud workspace
FROM CONVERSATION TO SOMETHING REAL Interactive demo · sample project
Your FellowBy workspacePreview only
Team & chatStudio booking site

YOUR PROJECT CONVERSATION

Start with a conversation.

MilesProject lead
You

I need a booking site for my design studio. Show our services, let people request a time, and keep the design calm.

Miles Project direction

I’ll turn that into a focused plan. You can review the scope and team before any work starts.

Your proposed planPages, booking flow and independent review
Your next idea starts here…Type or dictate. Review before sending.
A conversation becomes a direction.

Tell Miles what you want to make. Review the proposed next step before work begins.

Your serverYour AI accountsYour decisions

A REAL PROJECT, FROM THE FIRST BRIEF

See the work
come together.

Follow Cedar Studio from a conversation to a working website. Real screens, real interactions, and work you can inspect.

3 MINUTES · MUSIC & ON-SCREEN EXPLANATIONS

A fictional workshop website built in FellowBy. Waiting time is condensed. The demo sends no bookings or enquiries.

Read the walkthrough

The video has instrumental music and on-screen explanations, with no spoken narration.

  1. Start with the outcome. Describe the workshop website to Miles, keep the reference image with the conversation, and review the saved brief and requirements.
  2. Agree on the work. Meet Daniel, the builder, and Emma, the independent reviewer. Follow assignments and inspect an exact permission request before changes run.
  3. Try the result. Filter the real Cedar Studio workshops, open a class, check form validation and make a temporary reservation with fictional details. Nothing is booked, emailed or stored.
  4. Keep the context. Explore the Knowledge graph, specialist skills, account connection options and usage estimates.
  5. Review the evidence. Read the saved independent ticket report. Final assembly review and your acceptance are separate steps; work is never published automatically.

FellowBy runs on your own server. Your chosen AI provider processes the context you send to it.

MEET YOUR FELLOWS

Different strengths.
One shared direction.

Miles brings the plan together. Specialists bring the right skills. You review the scope and choose when the work is ready.

Meet the nine standard roles. Your project uses the specialists it needs, with room to create your own.

See the team at work
DIRECTION

Miles

Project lead & orchestrator

Turns your brief into a proposed plan and coordinates the team’s next steps.

ENGINEERING

Daniel

Application developer

Builds focused application features and supplies source files and test evidence.

QUALITY

Emma

Independent quality reviewer

Checks saved work against approved requirements and flags gaps before handover.

BRAND

Olivia

Brand designer

Creates an original identity, brand guide and reusable design tokens.

UX / UI

Ethan

Product designer

Designs clear, responsive interfaces, including journeys, layouts and interaction states.

CONTENT

Sophie

Content designer

Writes clear product, landing-page and onboarding copy in your brand voice.

RESEARCH

Grace

UX researcher

Explores user journeys and product questions with findings you can trace to sources.

SECURITY

James

Security reviewer

Reviews authentication, permissions, input handling and sensitive data paths.

ACCESSIBILITY

Charlotte

Accessibility reviewer

Checks keyboard use, readable layouts and support for assistive technology.

Hover, focus or tap a tile to meet your Fellow.

LESS CHASING. MORE CONTEXT.

Keep the thread
through the whole project.

A message, a reference, a saved result.
Each one belongs with the work it helps move forward.

YOUR PROJECT CONVERSATION Example
YOUR DRAFT

Let’s make the booking action
easier to find.

Speak. Edit. Then send.

Say it your way.

Type a message or dictate into an editable draft. Local voice transcription keeps audio temporary. Nothing is sent to the agent until you choose to send.

SAVED WITH THE PROJECT Example
The agreed planScope and assignments
Reviewed
Saved work & notesContext for the next handoff
Saved
Independent reviewTied to the work you inspect
For your look

See what you’re saying yes to.

Inspect the output alongside its review. Acceptance belongs to the reviewed version, with feedback kept in the same project when it needs another pass.

A WORKSPACE OF YOUR OWN

Your workspace.
Your infrastructure.

FellowBy runs on your own computer or Linux VM. Your team uses it in the browser, with no FellowBy-hosted cloud workspace. You control project storage, access and backups; selected instructions and context go to your chosen AI provider.

See what you’ll need
Your browserYour team directs the work
Your FellowBy workspaceOn your computer or server
ConversationsPlans & filesReviews & evidence
Your chosen AI providerSelected instructions & context
Provider terms and usage costs apply

FOLLOW YOUR DATA

Your work lives in your installation.

Conversations, plans, saved files and review evidence stay in your FellowBy workspace on your computer or server. Your administrator controls access, updates and backups.

Clear choices.
Clear boundaries.

Your project workspace
Project records and saved work live in your installation. Selected context is processed by your chosen AI provider.
Registration, kept separate
Name and email are required during setup. They are separate from optional installation counts and usage sharing.
Usage sharing is your call
Off by default. If enabled, it excludes names, emails, voice audio and project content.
Your environment, your care
Your administrator manages HTTPS, access, updates and backups. Review AI work before applying or publishing it.

START WITH SOMETHING FOCUSED

Make room for
your next good idea.

For founders, studios and small teams who want to direct AI work from one shared workspace.

EARLY BETA · NON-CRITICAL PROJECTS

A first product experience

Explore a scoped site or prototype with a clear outcome and review before handover.

A focused next release

Turn a specific requirement into proposed work, reviewable changes and a clear next decision.

A team experiment

Try a different way to plan, assign and inspect AI work in an environment you control.

Start small. Beta workflows have limits and results need human review. Support for arbitrary existing codebases is not yet a general-purpose workflow.

YOUR COMPUTER. YOUR SERVER. YOUR CALL.

Your workspace.
On your terms.

FREE BETANo activation key.
Hosting and AI costs are separate.

Install once from the command line. Then plan, chat and review in your browser.

Choose where FellowBy will run.

Download and install without a browser sign-in or activation key. Your workspace stays on your computer or server. Internet access and your own AI account are needed.

Install on an Apple Silicon Mac EARLY BETA

A local workspace on your M-series Mac. Your files stay on your computer.

Apple M series · 6 GiB for Docker · 20 GiB free

  1. Get your Mac ready.

    Use an Apple Silicon Mac (M series) with at least 20 GiB of disk space free. The command below installs missing Homebrew, Python, Minisign and Docker Desktop, then starts Docker for you. Existing tools are reused.

    This build is for Apple Silicon. Intel Mac support has not been verified.

  2. Copy. Paste. Install.

    Open Terminal and run this one command. Follow any Mac permission or Homebrew prompts, and review Docker’s first-use agreement if shown. The installer opens Docker automatically and retries if its engine does not respond, then downloads about 3.1 GB, verifies the signed package and starts FellowBy.

    If Docker needs more memory, setup asks you to allocate at least 6 GiB in Docker’s settings and waits while you apply it. No download sign-in or activation key is needed.

MAC TERMINAL
curl --fail --silent --show-error --proto '=https' https://get.fellowby.com/install-mac.sh --output fellowby-install.sh &&
printf '%s\n' 'c1d9e719883bc80780ca08084c67637d9ab667a98d800579b0e3ae888f84ef39  fellowby-install.sh' | shasum -a 256 -c - &&
/bin/bash fellowby-install.sh --directory "$HOME/FellowBy"

Make the workspace yours.

When the installer is ready, open http://localhost:28794 on this Mac. Copy the Server ownership code shown in Terminal, then create your administrator login. Enter your name and email during setup and connect your AI account in FellowBy’s connection settings.

Usage sharing is optional and starts off. Registration is kept separate. Voice audio is temporary and stays local.

The CLI downloads and verifies the complete signed package, then starts your local workspace. Open the browser only after installation to create your account and connect your AI provider. The physical Mac trial remains the final compatibility check.

Install on a Linux VM

A shared workspace on infrastructure you control. Your team connects through HTTPS.

Linux AMD64 · AppArmor · Two HTTPS origins

1. Prepare your server

Use Debian 13 or Ubuntu 24.04 on an AMD64 VM. Replace both example domains with your configured HTTPS addresses, then run this command as a sudo administrator. It installs missing Docker Engine, Compose v2, Python, Minisign and AppArmor, starts Docker, and downloads the signed FellowBy package. No download sign-in or activation key is needed.

Before you begin

Environment
Debian 13 or Ubuntu 24.04, AMD64. Recommended: 4 CPU cores, 8 GiB RAM and at least 20 GiB free.
HTTPS
Two distinct HTTPS origins: one for the workspace and one for project previews. Configure DNS and the proxy before installation.
Download tools
The command installs missing Docker Engine, Compose v2, Python, Minisign and AppArmor from signed system repositories, starts Docker, verifies the signed FellowBy package and prints the setup code. Existing tools are reused.

You control DNS and your HTTPS proxy. If your HTTPS proxy is on another private machine, add --trusted-proxy 192.0.2.10 (replace this example with its private IP) to enable the restricted VM relay on ports 8080 and 8081; FellowBy controls preview access through your signed-in session. Avoid adding a second password prompt on the preview domain. Connect your AI account after browser setup.

2. Run the installer

LINUX VM TERMINAL
sudo apt-get update && sudo apt-get install --no-remove --no-upgrade -y ca-certificates curl &&
curl --fail --silent --show-error --proto '=https' https://get.fellowby.com/install-linux.sh --output fellowby-install.sh &&
printf '%s\n' 'd081bc6556dba3c630688d1e614bfd6bbecfd2caf4e3b13011958a4a6d7d2c6d  fellowby-install.sh' | sha256sum --check --strict &&
bash fellowby-install.sh \
  --directory /opt/fellowby \
  --name fellowby-beta-team \
  --origin https://studio.example.test \
  --preview-origin https://preview.example.test

3. Open your workspace

Keep the setup code printed in this Terminal. Open your configured workspace address, enter that code, then create an administrator account and register your name and email. Optional reporting and usage sharing start off. Connect your AI account after setup.

Prefer a manual download? Download the Linux ZIP. The signed manifest and file verification requirements still apply.

Install on a Windows PC EARLY BETA · X64

A local workspace, installed from PowerShell and powered by Docker Desktop.

Windows x64 · Linux containers · 20 GiB free

1. Get your computer ready

Use Windows x64 with Docker Desktop running in Linux containers mode. Allocate at least 6 GiB to Docker and keep 20 GiB of disk space free. Install Python 3.11 or newer and Minisign from their official sources. Make sure both tools are available in PowerShell.

2. Sign in and install

Open PowerShell in an empty folder and run the command below. It downloads the signed package and installs FellowBy without a browser sign-in. It works in Windows PowerShell 5.1 and PowerShell 7.

WINDOWS POWERSHELL
$ErrorActionPreference = 'Stop'
Invoke-WebRequest -UseBasicParsing -Uri https://get.fellowby.com/install-desktop-x64.py -OutFile fellowby-install.py
py -3 -I -c "import hashlib, pathlib, sys; sys.exit(0 if hashlib.sha256(pathlib.Path('fellowby-install.py').read_bytes()).hexdigest() == 'fd96eb45718c46b279926700371fb53e25d7105c4668a11241e1a448d3af2ebd' else 1)"
if ($LASTEXITCODE -ne 0) { throw 'Installer checksum failed.' }
py -3 .\fellowby-install.py --directory "$HOME\FellowBy"

3. Open your workspace

Keep PowerShell open until the installer reports ready. The cold Windows test took about ten minutes, including download and first startup; timing varies. Then open http://localhost:28794. Copy the Server ownership code shown in PowerShell into browser setup. Finish registration and connect your AI account in the browser. Usage sharing starts off.

Fresh Windows Docker Desktop setup, isolated execution, restart and native voice checks have passed. The public CLI delivery is tested separately. Windows ARM has not been qualified.

Install on a Linux PC EARLY BETA · X64

Your own workspace with a local Docker Engine or Docker Desktop.

Linux x64 · Docker Compose · 20 GiB free

1. Get your computer ready

For a local Linux x64 computer, use a local Docker Engine with Compose or Docker Desktop. You need Python 3.11 or newer, curl and Minisign on your PATH, at least 6 GiB available to Docker and 20 GiB of free storage. Install these from their official package sources. The desktop package keeps its execution sandbox inside a dedicated Linux guest; it does not turn off security profiles on your PC.

2. Sign in and install

Run this in a terminal on your PC. The command downloads and verifies the signed package without a browser sign-in. This local workspace listens on loopback; use the Linux VM tab for a shared remote server.

LINUX PC TERMINAL
curl --fail --silent --show-error https://get.fellowby.com/install-desktop-x64.py --output fellowby-install.py &&
printf '%s\n' 'fd96eb45718c46b279926700371fb53e25d7105c4668a11241e1a448d3af2ebd  fellowby-install.py' | sha256sum --check --strict &&
python3 fellowby-install.py --directory "$HOME/FellowBy"

3. Open your workspace

Open http://localhost:28794 when ready. Copy the Server ownership code shown in Terminal, finish registration and connect an AI account. Optional usage sharing is off.

The signed x64 desktop package is available for early beta testing. A separate physical Linux PC run and distribution-specific lifecycle checks are still pending; this is not a promise of support for every Linux distribution.

Install on an Intel Mac HARDWARE TEST PENDING

The x64 desktop package with Mac Terminal and compatible Docker Desktop.

Intel Mac · 6 GiB for Docker · 20 GiB free

1. Get your computer ready

Use an Intel Mac with a supported macOS version and at least 20 GiB free. This command installs missing Homebrew, Python, Minisign and Docker Desktop, starts Docker and continues with FellowBy. Complete any Mac permission, Homebrew or Docker agreement prompts yourself. If asked, allocate at least 6 GiB to Docker. Existing tools are reused; physical Intel Mac qualification is still pending.

INTEL MAC TERMINAL
curl --fail --silent --show-error --proto '=https' https://get.fellowby.com/install-mac.sh --output fellowby-install.sh &&
printf '%s\n' 'c1d9e719883bc80780ca08084c67637d9ab667a98d800579b0e3ae888f84ef39  fellowby-install.sh' | shasum -a 256 -c - &&
/bin/bash fellowby-install.sh --directory "$HOME/FellowBy"

3. Open your workspace

2. Sign in and install

The installer detects Docker’s architecture and selects the x64 package. When ready, open http://localhost:28794, copy the ownership code shown in Terminal and complete browser setup. No activation key or AI account is bundled.

Intel Mac execution and performance have not been tested on physical hardware. This is an experimental beta path; retain your existing installations and start with a new folder.

Already have Docker Desktop? CHOOSE YOUR HOST

Reuse your Docker engine, then choose the installation command for your computer.

Your existing engine · Linux containers · 6 GiB allocated

Docker Desktop is the engine used by the computer installations, rather than an extra operating system. Start it, use Linux containers and allocate at least 6 GiB. Then choose your host’s command below. The installer selects the signed image, creates isolated storage and starts the local workspace.

Check that Docker and Compose respond before installing:

docker version
docker compose version

Your workspace opens at http://localhost:28794. No privileged container, host Docker socket mount or disabled security profile is required. Internet access and a supported AI account are still needed.

Optional install measurements

Help us improve setup and understand which campaigns work. Sharing starts off, and installation works without it.

What gets shared?

Your install environment, method, release, setup stages, broad timing or a fixed failure category, plus a public campaign reference. Sent to PostHog in the US. No name, email, browser identity, project data, command text or local paths. Browser setup asks separately before sharing its completed-setup report.

For manual ZIP setup, copy the optional campaign reference below. This is for measurements; it is not an activation key.

No campaign code included.

Campaign handoff is off.

Optional Google ad measurement

A separate choice, off by default. Installation works without it.

What gets shared?

After a recognized Google ad visit, this choice lets FellowBy send Google your ad-click ID, successful CLI install time and a deduplication code. No names, emails, project content, command text or local paths. Click IDs stay out of PostHog.

The copied command contains a temporary random completion token, never your click ID. The browser remembers this choice for up to 30 days. You can untick it here to stop future reports, including from earlier copied commands. Data already imported by Google is not recalled. FellowBy deletes its click record after 30 days. Connection IPs are processed in transit, not saved in these records or forwarded to Google.

Ad measurement is off.

A FEW USEFUL ANSWERS

Before we
make a start.

Still have a question?
Say hello

Is the beta free?

Yes. The beta requires no paid FellowBy licence or activation key. You cover your own hosting, any applicable Docker costs and AI provider usage. Future paid features or plans may be introduced.

Can anyone download it now?

Yes. Installer downloads are public and free during the beta. Choose your environment and run its command; no download account, browser sign-in or activation key is required. You enter your name and email during first setup. Optional usage sharing starts off.

Do I need to use the terminal every day?

No. Use your computer’s terminal or PowerShell, or SSH into your Linux VM, for installation and maintenance. Your team directs the project, chats with Fellows and reviews the work in the browser.

Can I use my existing AI account?

FellowBy supports provider connections including Codex and Claude Code. Connect a supported account on your installation, then choose the provider and model for your Fellows. Availability, limits and charges follow the provider’s terms.

What information is shared with FellowBy?

Setup requires your name and email for registration. Installation counts and usage sharing are separate, optional controls, off by default. Optional analytics excludes names, emails, voice audio and project content. Your chosen AI provider still processes the instructions and context you send to it.

How does voice typing work?

whisper.cpp transcribes speech locally into an editable message draft. You review it before sending to the agent. Audio is temporary and never included in analytics.

Who handles hosting, updates and backups?

Your installation administrator does. FellowBy runs in your environment, so your team manages the computer or VM, access, HTTPS for remote access, upgrades and recovery copies. Start with non-critical work and review AI output before applying it.

GOOD WORK, MADE TOGETHER

Your next idea
deserves company.

Start with a conversation. Keep the direction.