The workspace your team keeps building
Chat, tasks, docs and huddles for your team and its agents.
Ask AI to extend it with full-stack apps.
$48.2k
+6.4%
61%
+3.2 pts
4 playing
For lean teams that want the most from AI
One shared context, with the AI inside every surface rather than bolted on the side.
All features18–34 hours a week, per person, back
That’s thousands of dollars saved, a week.
Our estimate, not a measurement - and the top of each range is a week that is mostly coordination already. Halve every number and the week still comes back.
17.9–34.4 hThe workspace
Channels and one board, all of it in one context.
Multiplayer
Claude Code, Cursor or Codex join over MCP. Same board.
Coding Agents
A cloud agent works your repo and opens the pull request.
Memory and skills
Everything it does becomes context. It proposes new instructions as pull requests.
Voice-to-voice
Press ⌘O and talk. The workspace is already in context.
Deep Research
Searches the live web for what the workspace doesn’t hold.
Commands
Save any prompt as a command. Run it, or schedule it.
Ambience
Fact-checks comments and flags decisions that clash.
Decisive for your teams, Slack for your organization
Nothing here asks you to move the organization off Slack. The org keeps talking where it already talks - the teams that ship get a smaller room, with the work and the agents inside it.
Our coding agent solves your tasks
Assign one. It works in your repository on its own branch and opens a pull request you review.
Connect your local agent, see live who’s on what, and get warned before you duplicate work
Run it in your repo root, and whatever you code with - Claude Code, Cursor, Codex - reads the tasks, moves the board and answers teammates over MCP.
An engine built so the AI can read every detail - at a fraction of the cost
Not search, not retrieval, not a summary - the AI reads the workspace itself, so it already has whatever your question needs. Caching between turns is what keeps that cheap enough to do every time.
Radical choices made it possible
No DMs, no private channels, no seventh tab.
Read whole, not retrieved
No search, no ranking, no guessing which snippets matter.
Always live
What changed a second ago is already in.
Caching makes it extremely efficient
Everything unchanged is read from cache, not paid for twice.
Custom apps
@AI ship us a retention metrics dashboard
Ask on the Apps tab and your team has it by the end of the sentence - an invoice tool, a rota, whatever your team keeps doing by hand. Decisive builds it, hosts it on its own address, and keeps changing it with you in a thread - real npm packages, its own database, no repository to set up.
Signup funnel
PostHog · 7 days
Support queue
Intercom · 12 open
Net revenue
Stripe · this month
$48.2k
+6.4%
Standup board
Tasks · 3 of 5 in
Friday quiz
4 playing
Retainer invoices
March · $18,400
Decisive
Emma mentioned you in Chat
the theme flickers on load - take a look?
Import your data
One field. Paste your data in, or drop a file. The AI reads it, works out what came from where, and shows you an import plan. Nothing is written until you confirm.
Export whenever
One button in Settings downloads the whole workspace as a .zip - every task,
discussion, page, comment and message, as complete records in workspace.json plus readable markdown. No plan gate, no request form.
Free forever
Up to 10 seats, 1,000 AI credits and one app. Everything on this page except voice.
Billed annually
Adds voice, unlimited apps and unlimited seats.
Billed annually
Coming soon - SSO, invoice billing and cross-team collaboration.
Frequently Asked Questions
What is Decisive, exactly?
One workspace instead of separate chat, task, doc and meeting tools. It connects to your GitHub repository, and an AI teammate works inside it - answering in chat, triaging tasks, writing docs, opening pull requests. Your own terminal agent joins over MCP.
Who is it for?
Software teams that ship from a GitHub repository - whether that is a founding team of three or an established engineering team that already has its tools. A workspace has channels, but every one of them is open to the whole team and the AI reads them together, which is what keeps a single context it can actually use; larger organizations run a workspace per team.
How does the AI open pull requests?
Assign a task to @AI or ask in chat. A coding agent clones your repository in a cloud sandbox, works on its own branch and opens a pull request. Nothing lands without your review.
What context does the AI actually have?
The whole workspace: chat, tasks and comments, discussions, docs, huddle summaries, your labels, statuses and sagas, and who is on the team. Recent commits join once you connect the repository in Settings → GitHub. Assembled fresh on every request, never pulled from a stale index.
Can we bring our data from Linear, Notion or Trello?
Yes. Paste your data into the Migration assistant, or drop a file in, and the AI shows you an import plan before anything is written. You are never asked which tool it came from. Files need to be text: CSV, JSON, Markdown or a .zip.
Can we talk to someone first?
Yes - book a call at wemachines.com/book-a-call and pick a time that suits you. Nothing needs a call, though: the free plan is the full workspace for up to 10 people, with no card and no demo to sit through.
What does it cost?
Free for up to 10 seats, with the MCP server and one app included. Start-up is $10 per seat per month billed annually, and adds voice and unlimited apps. Every workspace starts with 1,000 AI credits, and more come as monthly packs.
Which model powers it?
The built-in AI runs on frontier models we keep current, so there is nothing to configure. Coding runs, app builds and your agents can be pointed at any model in the catalog - hundreds, across every major maker, one bill in credits.
Can I use my personal Claude account?
Yes - two ways now. Install Decisive in your repo and your own Claude Code syncs with the team in real time, on your own plan: it reads the workspace, moves tasks and answers colleagues without leaving the terminal. And in the Mac app you can hand a task straight to your own Claude Code - the coding agent runs on your machine, on your subscription, for no credits. That run is deliberate and contained: you start it, it works in an isolated git worktree beside your checkout, and a guardrail layer blocks dangerous commands outright while the rest runs unattended, unless you set the run to ask you. What we still refuse is the opposite shape - the platform reaching out to an always-on agent on your laptop, where everything landing in the workspace becomes an instruction to an agent holding your shell, your files and your keys. Cloud agents are there too, for when you are away from your machine.
Can we run the AI on our own infrastructure?
Partly today, and fully is where we are headed. The Mac app already runs a task’s coding agent on your own machine through your own Claude Code, so that work never leaves your laptop. Running the whole platform - the model and the agents - on hardware you control is a direction we are committed to; no date yet, and we will say so here when there is one.
Is there a desktop app?
Yes - a native Mac app: 7.2 MB, universal for Apple Silicon and Intel, and it keeps itself current. Windows and Linux are in progress.
Is there a mobile app?
Yes - install it from your phone’s browser and add it to the home screen. It runs full screen, with push notifications for mentions, replies and tasks moving. On iPhone that home-screen step is what switches push on. A native app is in progress.
Is Decisive open source?
No, and that is a deliberate decision rather than a stage we are at. Decisive holds your team’s private work - chat, documents, task history and access to the repository you connect - and publishing every line that guards it hands anyone probing for a way in the same map we work from. We are not willing to take that risk with other people’s data.
Transparency about what happens to your data does not require public source: the security page sets out how workspaces are isolated, what reaches an AI provider and how the coding agent is sandboxed, the privacy policy names every sub-processor that can touch it, and anything still open gets a direct answer - [email protected].
The AI already clocked in.
Free for up to 10 people, with 1,000 AI credits included. No credit card, no sales call, no onboarding deck.