Get access

Features

Everything your team and its machines can do in WeMachines, in one list.

Work together

Team chat

Talk with your team and AI in the same channels. Mention a machine to ask a question, create a task or update a doc.

Guide: Talk to your machines
  • Channels A workspace starts with #general and a voice room. Admins add more, and every channel is open to the whole team.
  • Mention a machine Type @ and a teammate’s name, human or AI. The machine answers in the same channel and can act on what it read.
  • Turn talk into work Ask the AI to file what was agreed. Tag a task, doc or app with # and it shows as a live chip in the message.
  • Threads Reply in a thread to keep a side conversation out of the channel.
  • Whispers and machines Whisper a person one-to-one, end-to-end encrypted, and only they are notified. Each AI teammate has its own chat the whole team shares.
  • Machines that join in Call a machine by name and it answers without a tag. Set one to Joins in and it can step into talk that falls inside its brief.
  • Files and previews Up to ten files a message, video that plays inline, and previews for the links you paste.
  • Save for later Set a reminder on any message, and mute a channel just for yourself.
  • Shared channels Open a room with another company through a single-use link. Only messages cross, and the AI stays out.

Good to know

  • Every channel is readable by the whole team. Whispers are end-to-end encrypted between the two people in them, and only the recipient is notified. There are no private channels.
  • Chat is kept. A thread quiet for 14 days moves into the grey zone: greyed out, found by ⌘K, and left out of what machines read by default. A lasting decision belongs in a doc or a task, and the AI can put it there.

Tasks

Keep tasks in a list or board. Ask AI to create and update them from chat. Connect GitHub to move tasks as pull requests open and merge.

Guide: Bring your work over
  • Tasks from chat Ask in chat, or add a task yourself. The AI checks for an existing open task before creating one.
  • List or board Group by status, topic, assignee, priority or label, as a list or as a board.
  • Moved by your code Put the task’s code in a branch or a pull request, and the repo’s workflow rules move the task as the PR opens and merges.
  • Your own statuses Name and colour your columns. Each one is a backlog, not started, in progress, completed or cancelled kind.
  • Codes you can type Every task has a code like T-42, so chat, commits and pull requests can point at it.
  • The pull request, live A linked PR shows its state on the task: draft, open, merged, changes requested or checks failing.
  • Update tasks in batches Ask the AI to triage the backlog and it works through the whole board in batches.
  • Code with a Machine Press Code with a Machine on any task. It works on a branch and opens a pull request.
  • Share one, read-only A public link shows a task’s code, title, status and pull request, never its body or comments.

Good to know

  • Tasks have no due dates or estimates. Order and priority carry the plan.
  • A finished task is kept. After 14 days in a completed status it moves into the grey zone: out of the list until you ask for it, and still found by search. Reopening it brings it back.

Docs

Write plans, specs and decisions together. Select a passage and ask the AI to edit it. You can restore the version it replaced.

Guide: Talk to your machines
  • Write together Pages edit live for everyone, with codes like P-7 and blocks you can toggle and drag.
  • Ask AI for a change The AI composer sits at the foot of the page. Select a passage to keep the edit to it.
  • Restore a version Each page keeps the last 20 versions replaced by AI edits. You can restore any of them.
  • Inline comments Select text and comment. Each thread has its own link, and a machine you mention sees the passage it is on.
  • Charts from a table Add a /chart block and type into its grid, or paste a spreadsheet straight in.
  • What the AI reads Set each page to Always in context, Title only or Hidden from Machines.
  • Who has read it A strip on the page shows who has read it, so a decision doesn’t rest on a guess.
  • Find and outline ⌘F finds on the page, and an outline appears once a page has three headings.
  • Pages that last Docs don’t expire. They are where the decisions from chat should end up.

Good to know

  • Versions keep what an AI edit replaced. Edits made by people are not versioned.
  • Machines look at attached pictures and read attached PDFs, PowerPoint, Word and Excel files. A zip or a clip, they know only by name.

Files

Ask a machine for a pitch deck, a report or a budget and get a finished file. Every file your team makes or shares is kept in one place.

  • Decks, documents and sheets A machine writes the words and the numbers, and the design is done for it: slides with charts and big numbers, documents with headings and tables, sheets with formulas and totals.
  • Download for any app A deck downloads as PowerPoint or PDF, a document as Word, PDF or Markdown, and a sheet as Excel or CSV.
  • Ask for a change Ask the machine to add a slide or rewrite a section. Each change is a new version, and you can restore an older one.
  • Present from the browser Open a deck and press Present to show it full screen.
  • One place for every file What machines make, what you upload and what was shared in chat, newest first. Narrow it by kind or find a file by name.
  • Storage for the team Start-up includes 50 GB of file storage per seat and Grow 100 GB per seat, pooled across the team. Add more in packs of 100 GB.

Good to know

  • Files are for your team only. There are no public links.
  • Office files you upload download to open. The preview shows what a machine made, pictures, PDFs, clips and text.
  • Machines read the files here when a question needs one. Keep one always in their context, or hide it from them.

Huddles

Join a voice room from the channel list. Talk, turn on video or share your screen. Calls are transcribed, so you get a summary and transcript after the call.

  • Drop-in voice rooms Join a voice room from the channel list. Several can be live at once, each with its own chat.
  • Video and screen share Video, screen share and reactions. Move around the app and the call follows you in a corner window.
  • Summary and transcript Calls are transcribed while mics are on, and anyone in the call can switch it off. After it ends, the AI posts a summary and the full transcript.
  • Transcripts, not recordings Audio and video are never recorded. Transcription is per call, and anyone in it can switch it on or off.
  • A summary on Transcripts Your primary machine writes the summary on the Transcripts page, and says which room the call was in.
  • On every device Join from your laptop and your phone at the same time.
  • Guests by link An admin can open a room to guests. They knock without an account, and someone in the call lets them in.
  • Guests see the call A guest isn’t transcribed and sees none of the workspace.
  • A chat of its own Every room keeps its own chat for links and notes during the call.

Good to know

  • AI teammates don’t join calls. They write the call up after it ends.
  • Transcripts and summaries are kept, and move into the grey zone 14 days after the call started: greyed out, shown on Transcripts when you ask, and still found by search.

Polls

Type /poll in a channel and add up to ten options. Teammates vote on the card, and everyone can see the results.

  • /poll in any channel Type /poll with a question and its options, or pick Poll from the / menu.
  • One pick or several Vote for every option you could live with, and click again to take a vote back.
  • Live results Results update live, with each voter’s face on their pick. There is nothing to close.
  • Two to ten options Enough for a real choice, few enough to read at a glance.
  • Faces on the votes Everyone sees who picked what, so a vote can start the next conversation.
  • The AI can ask Ask the AI to put a question to the team and it posts the poll. It never votes.
  • On a schedule A routine can post a poll for you, like a Friday check-in.
  • In search Polls come up in ⌘K with the rest of the team’s work.

Good to know

  • Votes aren’t anonymous, and a poll has no deadline.
  • Polls live in chat, so a poll moves into the grey zone when its thread has been quiet for 14 days. Put the outcome in a doc or a task.

Inbox

See new tasks, docs, comments and messages that involve you. Open an item to read it, or filter the list to show unread updates.

  • What involves you New tasks, docs, comments, thread replies and the chat that involves you land here as they happen.
  • Unread stands out Open an item to go straight to it. Unread items stand out until you have seen them.
  • Clear it in one go Mark everything read in one go, or keep only what is unread on screen.
  • One count everywhere The Inbox button, the app icon and the chat marks all count the same list.
  • You set the level Choose what lands for chat, tasks and docs in Settings. The default is what involves you.
  • Finished builds When an app you asked for is ready, it lands here.
  • Reminders Ask the AI to remind you on Friday at 3. It waits on the Reminders tab, then rings here and on your devices.
  • Shared channels Messages from a room you share with another company arrive here too.
  • By day Items are grouped by day, newest first.

Good to know

  • The Inbox lists what happened, newest first. It doesn’t rank items by how urgent they are.

Build and connect

Apps

Describe a tool your team needs. WeMachines builds and hosts it. It can read your tasks, docs, chat and repo. Ask for changes in a thread your team can see.

How to build an app
  • Built from a description Say what it is for, who will use it and where its data comes from.
  • Watch it build The thread says what the AI is doing in plain English, with screenshots, and anyone can press Stop.
  • A version for every change Ask for changes in the same thread. Every turn saves a version, and its icon on the Apps page opens it in a window.
  • Reads your workspace An app can read your tasks, docs, members, chat and connected repos. It can’t write to them.
  • Storage for app data An app that stores data can have its own API and database.
  • Keys stay private An admin adds a service’s key in the app’s settings, where nobody can read it back.
  • AI inside An app can use the team’s AI on credits, with no key to paste.
  • The model per turn Choose which model builds each change: Claude Opus or Sonnet, or GPT Astra or Sol. On the desktop app, a Local build runs on your own coding agent.
  • Export the app Read an app’s code and stored data in the app, or export the whole thing as a zip.

Good to know

  • Apps are in early access: the newest part of WeMachines, and the part that changes fastest.
  • Apps are made for your own team and open only for people logged in to your workspace. Publishing an app to the internet is switched off.
  • Start-up keeps up to 20 apps and Grow up to 50.

GitHub

Connect your repositories and link commits and pull requests to tasks. Set rules to update task status when code is pushed or a pull request opens or merges.

Guide: Connect your repo
  • Every repo you ship from Add every repo your team works in under Plugins → GitHub, each with a line on what it holds.
  • Task codes in your commits Put T-42 in a commit subject, a branch name or a pull request title, in any connected repo.
  • Workflow rules Each repo’s workflow rules move tasks forward when code is pushed, a PR opens or a PR merges.
  • Live pull request state The pill on a task shows draft, open, merged, changes requested, failing checks or conflicts.
  • Forward only Rules only move a task forward, so a late push never drags finished work back.
  • Reviews on the task Review comments show on the task. Press Address and a machine replies in each GitHub thread.
  • Code questions answered Ask where the retry logic lives and the AI picks the right repo, reads it and cites the paths.
  • A map of your repos Study works out what each repository is for and draws how they depend on each other.
  • The right repo, every run Choose the repository for a task. Machines use it for the branch and pull request.

Good to know

  • Repositories connect from GitHub only. GitLab and Bitbucket can’t link code to tasks or take a coding run.
  • Nothing merges on its own, and review comments are answered when a member presses Address.

Multiplayer

Connect Claude Code, Cursor or Codex to WeMachines using MCP. Your agents can read the team’s tasks and docs. Their current work appears on the board.

Guide: Connect your repo
  • Install in one line Run the installer in your repo, or add the server’s address to any MCP client.
  • Choose who it acts as Sign in as one of the workspace’s AI teammates to get its brief, skills and tools, or choose Read only.
  • Work in the open A live mark appears beside the person, the task moves to in progress, and the agent hears first if someone else is on it.
  • Free on every plan The MCP server costs nothing, and the agent runs on your own Claude or Codex plan, not on credits.
  • What the in-app AI can do Edit tasks and pages in place, and reach labels, statuses, apps, research and repos.
  • Briefed before it starts The agent learns whether a teammate’s AI is already on the task before it touches anything.
  • Signed for you What it posts in chat is signed by the AI teammate, for the person running it.
  • Writes where you allow Writes are blocked outside a repo set up for WeMachines, and auto mode can be on, ask or off.
  • The clients you use Claude Code, Cursor, VS Code, Codex, Windsurf, Zed, JetBrains and any other MCP client.

Good to know

  • WeMachines doesn’t supply the model for a connected agent. It runs on the plan you already pay for.

Plugins

Connect a service and let your machines use it. The directory holds more than 2,400 plugins, Supabase, Stripe, Sentry, Vercel and Jira among them.

Browse every plugin
  • Everyone uses their own account Connect your own account. When you ask a machine to use a plugin, it uses your access.
  • Or one account for the team An admin can connect a shared account for the team. It starts with read-only access.
  • Admins decide what machines can do Choose read only, ask first, or read and change. With ask first, you approve each change. Admins can switch off single actions, and a plugin only gains new ones when an admin connects it.
  • Encrypted logins and an audit log Keys and logins are stored encrypted. Changes made through plugins appear in the audit log.
  • Any MCP server Any service with an MCP server can be added by its address, and your machines treat it like any other plugin.
  • Built in GitHub, Google Calendar and Slack connect on every plan.

AI teammates

Machines

Add a machine for research, code review or release notes. Tell it the job and it names itself and writes its own brief. Your team can work with it in chat, tasks and docs.

Guide: Give each machine a job
  • Made by telling it the job Press + beside Machines in chat and tell it the job. It names itself and writes its own brief, or you set it up by hand.
  • Tell it how to work Ask it to keep its answers short, think harder or be more direct. It changes its own settings and keeps working that way.
  • Answers where you mention it It answers in chat, tasks and pages, and asks you back with a few options rather than guess.
  • A primary machine One teammate steps in where nobody is named: a call from a channel, call summaries, conflict checks and any routine that names no one else.
  • Its own voice and tools Pick one of six voices, the skills it may load and which of the team’s plugins it may use.
  • Effort Set how hard a machine thinks on a ruler from 0 to 25, or leave it on Auto, and set it for a single message from Effort under the message box. More effort means slower, more careful answers that use more credits.
  • A browser of its own Every cloud machine has a real browser that keeps its logins, so it can look something up on a site, click through an app or fill in a form.
  • Pictures and documents Ask for an illustration or a mockup and the picture lands on its reply. It also looks at screenshots and reads PDFs, decks, Word documents and spreadsheets.
  • Choose when it replies Mention the machine, address it by name or answer its question. You can also choose whether it joins conversations on its own.
  • It learns Correct it once and it keeps the lesson. Walk it through a procedure and it can save it as a skill, once you allow it.
  • Jobs across machines Ask for an outcome that spans several kinds of work, and it hands each step to the machine made for it.

Good to know

  • Machines work inside your workspace, with an email address and a browser of their own. They don’t join calls or shared channels, and they have no calendar of their own.
  • A local machine runs on the coding agent of whoever asks it, through the desktop app, with no credits spent. It can’t run routines or jobs.

Coding machines

Ask a machine to implement a task. It works on its own branch and opens a pull request for you to review. You can follow progress from the task.

  • Start it from anywhere From chat, from ⌘O, or with Code with a Machine on any task. Pick the model.
  • Asks before it steps out Reading, editing and running commands are approved from the start. For anything else, it asks.
  • A pull request to review The run ends in a pull request on its own branch, linked to the task.
  • Permission requests Allow once, allow for the run, or decline. A question left for ten minutes stops the run and saves its work.
  • A branch each Hand over several tasks and each run works on a branch of its own.
  • Local, on your plan On the desktop app, run it on one of 14 coding agents, like Claude Code, Codex or OpenCode, or on a model in Ollama, instead of credits.
  • Answers the review Press Address on a review and the machine works on the PR’s branch and replies in each thread.
  • No repo, still useful Without a repo, a run works with the web and a scratch shell and hands back a written result.
  • Spend under control Admins can switch off credit-paid runs and set a price ceiling on the models a run may use.

Good to know

  • Nothing merges on its own. A run on a repo ends in a pull request a person reviews.
  • Cloud runs use credits. Local runs on the desktop app use your own subscription.

Image generation

Ask a machine for a poster, illustration or mockup. Choose from image models such as GPT Image, Nano Banana and Seedream. Open or download the result from its reply.

  • Ask for any picture Mention a machine and say what you need: a poster, a product shot, an illustration, a logo idea or a mockup.
  • It checks, then draws If you left out the shape or the look, it asks one short question first. While it draws, the room shows it painting.
  • Full size or download The picture lands on its reply, shown large, to open full size or download like any file.
  • Choose an image model Each machine has an image model. The default is MAI-Image-2.6; you can choose another in its settings.
  • Change a picture Share a picture and ask for a change, like “make this one night-time”.
  • Your house style Set a machine’s usual shape, look and size once, and it stops asking.
  • In your apps An app’s AI can draw too, from the browser or the server, with its machine’s image model.
  • Paid with AI credits Pictures are paid from the team’s credits like any AI call, and show as Images in Settings → Usage.
  • A daily cap An admin can cap what the team spends on pictures in a day, or switch pictures off.

Good to know

  • A machine draws only when someone asks it to, never on its own. Routines and comments do not draw pictures.

Routines

Give a machine instructions and choose when to run them. Use a schedule, a webhook or another routine. It runs even when your team is offline.

Guide: Set up routines
  • Written like a message On the Routines page, under Machines, say what to do and where to report, or ask a machine in chat to set it up.
  • Choose a trigger A schedule, as often as every 15 minutes. A webhook address. Another routine. Or nothing, and run it by hand.
  • Reports where you want It reports in its machine’s room, a channel you name or a Docs page, or stays quiet when there is nothing to say.
  • Run it from chat Type / and a routine’s name to run it from chat.
  • Chains One routine can wake the next, up to four deep, and the ones in between can stay quiet.
  • Webhooks Point a service at a routine’s address, and the routine takes it from there.
  • Quiet runs A run can decide there is nothing worth posting and still count as a success.
  • Tasks a routine can run Research, file tasks, update docs, post polls, read the repo and start coding runs.
  • Stops when it breaks Three failures in a row switch a routine off, so a broken one doesn’t keep spending.

Good to know

  • Anyone on the team can create, change or run one. Viewers can only read.
  • Every run spends credits.

Voice command

Call a machine from any chat room, or press ⌘O. It can create tasks, update docs, manage connected calendar events and start coding runs. It answers out loud.

  • Call it from chat Press the voice key in a room’s message box, or ⌘O (Ctrl+O). In a machine’s own room you call that machine, anywhere else your primary machine.
  • Short answers, out loud Ask a question or ask for work. Answers stay short, because you are listening, not reading.
  • Work done while you talk Tasks, docs, polls, routines, apps and calendar events, made while you talk.
  • The same machine Each machine speaks in a voice of its own and keeps its brief and tone, so a blunt machine is blunt out loud too.
  • Just you Nobody else hears it, even when you are in a call.
  • Moves you around Ask for the board or a doc and it takes you there.
  • Starts coding runs Say “implement T-42” and a machine picks it up.
  • Sets up automation Ask for a routine or a new machine, and it makes one.
  • Your calendar It finds a time and creates, moves or cancels events on your connected calendar.

Good to know

  • Voice doesn’t run web research or read PDFs, and it isn’t for big batch jobs. Mention the machine in chat for those.
  • A call runs on one fast model, whatever model the machine uses in chat, so it starts talking within seconds.

Deep research

Ask a machine anything about your market. It splits the question into angles, researches them side by side and replies with a link to every source.

  • Ask in any channel Mention a machine with the question. When a plan or a strategy depends on the market, it researches without being asked.
  • Searches, reads, searches again A broad question becomes several, researched side by side by an agent that reads, then searches again on what it found.
  • Every source cited Every claim names its source with a link, and every figure carries its date.
  • Deep by default Every question gets the research agent, unless a quick lookup will do.
  • Side by side Up to five angles of one question are researched at the same time.
  • Recent when it matters News, launches and funding can be kept to the last twelve months.
  • On a schedule Put research in a routine, like a Monday competitor watch.
  • Into the work Ask for what it found as a doc or a task, and it writes one.
  • Paid at cost A deep question costs about 6 credits, a quick one under 1.

Good to know

  • The search provider keeps no data, so research works in a workspace set to zero-data-retention models only too.
  • It isn’t available by voice. Ask in chat.

Growth data

Ask a machine what your market posts, buys and complains about, which ads your competitors run, or where your next customers are. It pulls the numbers and tells you what to do next.

  • Ask for the numbers Name a competitor, an account, a hashtag, a market or a city. When growth turns on real numbers, a machine looks without being asked.
  • Live data, four platforms at once Profiles, posts, ads, reviews, search rankings, places, job posts and traffic, from up to four platforms at once.
  • The next move The answer ends with who to target, what to post and what to say, and turns into tasks when you ask.
  • Audiences Followers, top posts and engagement on Instagram, TikTok, X, LinkedIn and YouTube.
  • Conversations What people say about your market on Reddit, X and LinkedIn.
  • Competitors The ads they run on Meta, their reviews, launches, hiring and traffic.
  • Leads Businesses on Google Maps by category and city, with ratings and websites.
  • Search Who ranks on Google for the words your customers type.
  • On a schedule A weekly competitor watch or a daily lead list, posted to your team.

Good to know

  • Public data only: what anyone can see on the platform.
  • Paid in credits at what the data costs, usually 1 to 20 credits a lookup.
  • Off in a workspace set to zero-data-retention models only, and not available by voice.

Built on your work

Context

Machines use your team’s chat, tasks, docs, files and connected repo. Choose how each doc and file is shared with AI. Check Settings to see what each request includes.

Guide: Talk to your machines
  • It reads the workspace A request carries the team’s recent chat, its tasks and the docs you have put in context.
  • A setting on every doc Always in context, Title only or Hidden from Machines. A new page starts at Title only.
  • And on every file Machines see your files by name and open the one a question needs. Keep a file always in context, or hide it from them.
  • Check what it sent Settings → Context counts the tokens in every request and lists each item, biggest first.
  • About your team Your first conversation with the AI builds an About page for the team that rides along with every request.
  • Finished work stays light Done tasks go along as titles only.
  • Its own messages, optional Choose whether the AI’s own replies are part of what it reads.
  • Only this workspace The AI sees nothing from any other workspace.
  • Not for training Your team’s content isn’t used to train models.
  • The repo too Connect GitHub and machines answer code questions from the real source.

Good to know

  • Beyond the lessons a machine keeps, there is no separate memory store. Machines read what is current: what has moved into the grey zone after 14 days is left out by default, so decisions belong in docs.

Skills

Write your team’s instructions in SKILL.md files. Use them for code reviews, commits or releases. Machines load the relevant instructions when they need them.

Guide: Give each machine a job
  • A file in your repo Add a SKILL.md under .claude/skills or .wemachines/skills, with a one-line description.
  • Sync from the repo Anyone on the team presses Sync from repo in Settings and switches each skill on.
  • Loaded when it applies The one line tells a machine when a skill applies, and it reads the rest only then.
  • Per machine Narrow a machine to the skills its job needs.
  • Text, not code A skill is instructions. Nothing in it runs.
  • Free on every plan Skills cost nothing to add or sync.
  • Your terminal agents too A local agent that connects as a teammate gets that teammate’s skills.
  • Changed like code Skills live in the repo, so a change to one can go through review like anything else.

Good to know

  • A machine can save a new skill when you walk it through a procedure, but only once someone allows it. Skills in the repo are your team’s to edit.

Conflict checks

AI checks messages and comments against your team’s tasks and docs. If it finds a possible contradiction, it adds a card below the message for your team to review.

  • Every message checked Your primary machine compares a new message with related tasks and docs, and with general knowledge.
  • A card, if needed A contradiction gets a card naming what it disagrees with. Most messages get nothing.
  • Asks instead of guessing When two records disagree and the AI needs a decision, it asks with a card of options, and the answer is on the record for everyone.
  • Updates aren’t conflicts Announcing a change is an update. The check looks for claims that quietly disagree.
  • Skips the small stuff Very short messages and plain questions aren’t checked.
  • No duplicate tasks The AI checks for existing open tasks before creating one from chat, a routine or MCP.
  • A flag, not a gate The card sits under the message. The message is posted either way.
  • Comments too Comments on tasks and docs are checked the same way.
  • Built in It runs on WeMachines AI, as your primary machine.

Good to know

  • Only messages and comments are checked, not edits to a doc.

Charts

Ask a question about data in your apps, connected services or workspace. The AI can reply with an interactive chart. You can also add charts to docs, tasks and apps.

  • Ask in chat Ask something like “show us signups since the checklist shipped” and the answer is a chart.
  • From your own data It finds the numbers in an app, a connected service or the workspace, and uses the data it retrieves.
  • Check the values View the chart in the message. Hover to inspect values and check its source.
  • Twelve kinds Lines, bars, areas, donuts, funnels, scatters, heatmaps and more.
  • In docs Add a /chart block and type into its grid or paste a spreadsheet, with no AI needed.
  • From pasted text Paste notes into a chart block and press Chart it to let the AI shape them.
  • In pages and apps Pages the AI builds and apps your team ships draw with the same engine.
  • Sources on the figure A chart says what it shows, where the numbers came from and how many there are.
  • Readable by everyone Every chart has a legend, a data table behind it and keyboard support.

Good to know

  • The numbers come from your apps, your plugins and the workspace itself, never from memory.

Models and credits

Choose a model for each machine, coding run and app build. AI usage comes from your workspace’s shared credits. Set spending limits and review usage in Settings.

Guide: Spend credits well
  • A model per machine Each teammate has its own model, and moves onto another by itself when a provider is down.
  • Credits every month Every plan includes credits for each seat every month, shared by the workspace. Extra credits are bought from 500 at a time and never expire.
  • Spending limits Cap what a model may cost, decide whether machine tasks and apps may spend credits, and limit the turns in a run.
  • The newest models The picker reads the catalog live, so a model released today is there today and a retired one drops out.
  • Zero data retention Turn on zero-data-retention models only, and every picker offers just those. The built-in AI already runs on them.
  • See the spend Settings shows 30 days of spend by feature, member and model.
  • Your own plan, locally Coding runs on the desktop app and agents connected over MCP use your own subscription.
  • Replies pause, work doesn’t If credits run out, AI replies pause and everything else keeps working.
  • Top up automatically Enable automatic top-up to buy credits when the balance drops below your chosen threshold.

Good to know

  • A typical reply costs two to four credits.
  • Built-in features like summaries and conflict checks use models WeMachines picks and keeps current.
  • With zero-data-retention models only, machines still research the web and read links, but can’t read attached PDFs, look up growth data or use their own browser.

Try WeMachines with your team.

Something your team needs isn’t here? Book a call and ask us.

Get access Book a demo