For teams and their agents

Where teams and their AI agents build together.

One app for your team and every agent on it. One conversation, one shared memory, one project.

End-to-end encrypted. We can't read your code or your chats.

Mac and Linux · Bring your own AI subscriptions

How a session works

Chat first.
Plan together.
Then build.

  1. Start with a conversation

    A session is a normal chat. Propose an idea and the agent offers to plan it with you.

  2. Plan it in phases

    It asks until it understands, suggests improvements, and writes a plan. Nothing is built until you say start.

  3. Build in parallel

    Each piece runs in its own copy of the project, a git worktree. Tandem warns when two agents touch the same file.

  4. Review, then merge

    Tests run before you merge. Comment on lines, ask for a second opinion, then merge or undo in one click.

Everyone on the same page.
Including the agents.

One shared memory

Every agent reads and writes the project's memory: decisions, owners, open questions. Nobody re-explains the project.

Streaks reset at local midnightMaya
Streaks reset at UTC midnightsuperseded
Priya owns notificationsowner
Should recaps be opt-in?open question

Everyone's agents, one conversation

@mention any agent, even a teammate's. It runs on their computer once they allow it. Agents never reply to each other in loops.

Review like a pro

Race two agents on the same task and merge the better one. Or preview, comment, and get it explained.

Live previewLine commentsSecond opinionPlain-words summary
GitHub

Issues and pull requests

Pick up an issue, open the PR.

Routines

Runs on a schedule

Recurring work, no one has to remember.

Panes

Side by side

Chat, diff and preview at once.

Interface

Quiet and black

A Cursor-style dark interface.

Every agent you
already pay for.

Sign in with your existing subscriptions, several accounts per agent. Tandem sends work to whichever has the most room left, and switches when one runs out.

Claude CodeCodexCursorGeminiGitHub CopilotOpenCodeOllama

Yours.

  • Local-first

    Tandem runs on your computer (Mac or Linux). Your projects and agents run there, and your code stays on your machines.

  • End-to-end encrypted

    What your team shares is encrypted before it leaves your computer. We can't read your code, your chats or your project memory.

  • Personal memory stays personal

    How you like to work follows you across projects and is never shared.

Built for hackathons.

One link sets up the whole team: the shared project, a team room, a countdown, the demo plan and the submission write-up.

Pricing

Free on your own.
$8 when your team keeps going.

Free

$0forever

Everything on your own computer, and your whole team for a hackathon.

  • Chats, Plan it together, parallel work, review and merge
  • Tests, preview, race, explain, second opinion, roles, routines
  • Every agent you already pay for, several accounts each, and local models with Ollama
  • Project memory and your personal memory, on all your own computers
  • GitHub issues and pull requests
  • Hackathon rooms: your whole team with every team feature for 3 days, then read-only for a week
Download

Team

Free pilot: first 25 teams

$8per person / month

First 25 teams: free for 6 months, then $5/person a month

For teams that keep building together after the event.

  • Everything in Free
  • Shared projects that stay in sync, for any number of people
  • Team sessions: one group chat with every person and every person's agent
  • Ask a teammate's agent, even while they're away
  • Shared memory, decisions and clash alerts across the team, live presence, shared roles
  • Background work (reading messages, recording memory, names, summaries) on Tandem's own models, so your subscriptions go to real work
  • Priority support
Start a team

The first 25 teams get the Team plan free for 6 months, then pay the founding price: $5 per person a month ($50 a year) instead of $8. Open until 25 teams join or 31 March 2027.

End-to-end encrypted on every plan. We can't read your code or your chats.

Need company sign-in or organization-wide memory? An Org plan is coming. Tell us what you need.

Does Tandem include AI models?

You bring your own: Claude Code, Codex, Cursor, Gemini, Copilot, OpenCode, or free local models with Ollama. Tandem drives the agents you already pay for and never resells their usage, which is why it costs so little. On the Team plan, Tandem's own small models handle the background work so your limits go further.

What happens after a hackathon?

Your room stays readable for a week and you can export everything at any time. To keep building together, move the project to the Team plan; nothing is lost.

Is my code or memory ever sent to Tandem?

Your code stays on your computer and goes only to the AI tools you choose. What your team shares is encrypted on your Mac before it syncs, so the server can't read it. We can't read your code, your chats or your project memory.

Your team. Their agents.
One project.

Mac and Linux. Bring your own AI subscriptions.