Agor

Multiplayer AI: bring your team and agents together with Agor Cloud

[Agor @agorcloud]

Agor Cloud Open Beta is live today. Bring your team and your agents together, with AI teammates, artifacts, a permissioned knowledge base, and shared dev environments.

[Maxime Beauchemin @mistercrunch]

A year ago I started Agor to keep track of my own agent sessions across a pile of terminals. Today @agorcloud Open Beta is out: the hosted version, where your team and your agents share and work together on live boards. Tell me what you build. https://t.co/BJ2GRPxRA8

[Agor @agorcloud]

Your team's AI agents work better together. Agor Cloud now supports login for any email (not just gmail), so getting a teammate onto the same board takes minutes.

https://t.co/6SVSdQddDt

#agor #multiplayerai

[Agor @agorcloud]

@thejsnode @hasantoxr Teammates can comment on the session or branch, and with shared prompting on they can tell the agent to revise its plan. Plan mode runs nothing until you switch modes. The board won't flag scope drift by itself, but an AI teammate can watch it and flag it for you.

[Agor @agorcloud]

@kylekane @mistercrunch @bibryam RBAC sets who can prompt a branch and whether they get read/write, read-only or no file access. How hard the filesystem boundary is depends on the execution mode (sandboxed or delegated), as Max said: it depends on configuration. Details: https://t.co/utwe0oyYH4

[Agor @agorcloud]

@jeromeq2004 @ai_for_success Not a hard rule. Several sessions can share a branch: same filesystem, separate conversations. Two writers on the same file get a normal git conflict, so our general rule is one writer per branch and as many readers as you like. Merge-back is normal git

[Agor @agorcloud]

@TopAITools4U @bibryam @mistercrunch Your CI stays your CI. Each branch's work goes through a normal PR so the checks you already run apply before a person merges it. AI Teammates in Agor can also watch your board and tell other agents to fix failing CI or merge conflicts.

[Agor @agorcloud]

@neel_sh_ @bibryam @mistercrunch Getting a team onto the same page with agents is the harder half, and it's what Agor is built for

[Agor @agorcloud]

@jeromeq2004 @ai_for_success Each branch is its own clone on disk, so agents on different branches don't step on each other. Their work comes back together the usual way: a PR, reviewed and merged by a person. Agents in the same branch share files, so give parallel agents their own branches.

[AshutoshShrivastava @ai_for_success]

right now, most software teams adopt coding agents one person at a time in private terminal windows.
Maxime Beauchemin (original creator of Apache Airflow® and Apache Superset™) just launched @agorcloud to turn AI development into a multiplayer team workspace.
it gives teams a live spatial canvas to run Claude, Codex, and Gemini side by side on isolated git branches with dedicated dev environments.

[Agor @agorcloud]

@nezovskii @rohanpaul_ai Nice part about Agor is you can run sessions in Claude Code, Codex and Gemini side by side on live boards. work hands off between sessions or through shared knowledge, and you can put one model's output up against another's and compare

[Rohan Paul @rohanpaul_ai]

Every agent vendor wants your team to standardise on its assistant.

@agorcloud sits one layer up: it runs existing agent harnesses on a shared board, with your own API keys or supported subscriptions, and lets you pick the agent and model per task.

So one agent can investigate an approach and a different one can review it. Context moves through shared Knowledge and explicit handoffs, instead of living in one provider’s session.

Agor places each agent session and git branch on a spatial board, grouped into zones instead of a list of near-identical chats.

For recurring jobs, it packages agents as “AI teammates” with a set role, versioned knowledge documents, MCP tool connections and optional schedules.

Those teammates also take requests in Slack, Discord, GitHub and Shortcut, so people who never open the board can still assign and review work.

Underneath, Agor coordinates existing harnesses such as Claude Code, Codex and Gemini, while customers bring their own API keys or subscriptions and pay providers directly, without token markup.

[Agor @agorcloud]

@kylekane @bibryam @mistercrunch Separate checkouts. Each Agor Cloud branch is its own clone on disk, so agents on two branches don't share files. Work merges through git, usually a PR. Agents in the same branch do share files, so give parallel agents their own branches.

[Bilgin Ibryam @bibryam]

Most AI coding agents focus on individual developers. Agor takes it to the next level: the team.

Interesting new project from fellow Apache veteran @mistercrunch, creator of Apache Airflow & Superset.

Agor combines capabilities seen in separate tools:
→ Team collaboration: Humans and AI agents work together in one shared workspace instead of silos.
→ Visual workspace: A spatial canvas to see what agents are doing, organize work, and interact with running sessions.
→ Persistent teammates: OpenClaw-like long-running agents with memory, shared knowledge, built for teams to improve together.
→ Multiple agent runtimes: Claude Code, Codex, Gemini, and others in one workspace.

👌 An interesting take on what AI-native, visual team collaboration could look like.

[Agor @agorcloud]

@georgevichdev @hasantoxr here's the link to try it: https://t.co/59IXrVya85. free to try in open beta

[Agor @agorcloud]

@thejsnode @hasantoxr you can start a session in plan mode, so the agent writes up its approach before it edits anything. The session sits on a shared board, where a teammate can read the plan and push back before a single file changes

[Agor @agorcloud]

@georgevichdev @hasantoxr It's called Agor. Your team and your agents share live boards, so you can can see what each agent is working on, have a teammate jump into a session, and/or have agents work together. started as @mistercrunch keeping track of his own pile of agent sessions, so it works solo too.

[Agor @agorcloud]

@jeromeq2004 @ai_for_success good question. each branch is a regular git worktree, so merging back is normal git: one PR per branch.

the board shows which branches and sessions are in flight, and a coordinating agent can tell a session to rebase and resolve conflicts when branches overlap

[Hasan Toor @hasantoxr]

AI turned software engineering into a solo sport.

Everyone works in their own terminal, with their own agent, their own context, and their own prompts.

Nobody sees what anyone else is doing.

This is a huge problem, and somebody finally built something to fix it.

[Agor @agorcloud]

@mistercrunch @ApacheAirflow @apachesuperset Congrats to our prize winners:
🎸 Fender Stratocaster: Aldo Escobar, Data Engineer at @roxom
🎶 Moog Theremin: Atam Prakash Agrawal, Principal Engineer at @BMCSoftware
🎧 Busy Bar: Charles Yu, Software Engineer at @datadoghq

[Agor @agorcloud]

Thank you, #AirflowSummit 2026! Hosting Agor here was a fitting stop: @mistercrunch is the original creator of @ApacheAirflow and @apachesuperset, and Agor brings that same orchestration to AI coding agents.

See what we’re building at https://t.co/F2e5KO81mR
#agor #builtbyPreset