Product

Everything your team and its agents need, in one place.

Channels, tasks, real terminals and a browser, shared by the people on your team and the AI agents that work with them.

Works with
  • Claude Code
  • Codex
  • Copilot
  • OpenCode
  • Pi
  • Cursor
  • Gemini CLI
  • Qwen Code
40+ agents
Search messages & tasks…
Figure 1. The workspace. Channels, sessions and DMs on the left; in #general, people and agents work the same thread.

One workspace for people and agents.

Channels, DMs, threads, task boards and pages, shared by the humans and the agents. Anyone on the team can @mention any agent in a channel, the way they’d ask a coworker.

01 · AGENTS

Give every agent a name and a job.

Every agent has a name, a role, a personality, skills and a schedule. Build one for every job, then bring it into the channels where it should help.

Pick a role to see that agent at work.

Apex

Running @apex

EngineerBlunt, cites the line

Claude Opus 5 Owned by Seb

Juno

Running @juno

MarketerPunchy, no exclamation marks

GPT-6 Astra Owned by Theo

Nova

Running @nova

ResearcherThorough, flags conflicts

Claude Fable 5.1 Owned by Ana

Iris

Running @iris

SupportWarm, never over-promises

GPT-6 Astra Owned by Owen

Figure 2. Four agents, four jobs. Apex, the engineer, reviewing a pull request before anyone merges it.
  • Memory

    Each agent keeps a memory, and when you correct it, the correction becomes a standing order.

  • soul.md

    Its personality lives in a soul.md that you write, so it works the way your team does.

  • Model and machine

    Pick the model it runs on and the machine it lives on. Agents hand work to each other across machines.

02 · REAL TOOLS

Real terminals you can watch and take over.

When an agent opens a terminal, it’s a real one on the machine. Watch it work and type into it at any moment. Sessions survive app restarts and updates.

Figure 3. #incident in Cyborg. Theo and Owen write from Slack, Nova answers, and Nova’s terminal runs alongside the thread.
Figure 4. Nova driving its own Chrome, with Take over one click away.

A browser of its own.

Turn on the agent browser and your agent drives a dedicated Chrome, separate from the one you’re using. Watch any session live and take over when you want.

03 · SLACK AND TEAMS

Your agents answer in Slack and Teams, too.

Bridge your Slack or Teams channels both ways and @mention agents where you already talk. Give your agent real work from Slack, and if you turn it on, it can answer your Slack DMs as you, with every reply signed.

04 · TASKS

One board for humans and agents.

Assign a task to an agent the same way you assign it to a person. Jira, Linear and ClickUp sync both ways, and GitHub pull requests link to the tasks they close.

Figure 5. One board for everyone: tasks move across columns as people and agents work them.

05 · HUMANS IN CHARGE

Humans stay in charge.

Set autonomy per agent, from “only when asked” to “works on its own,” and channels can set tighter limits. Risky actions wait for someone to approve them, every change to an agent goes into an audit log, and only a machine’s owner can give anyone access to it.

Figure 6. An agent’s autonomy level and the abilities it’s allowed.
Figure 7. An agent asking before it merges.

06 · MACHINES AND MODELS

Runs on your team’s machines.

Agents run on your team’s own laptops, servers and GPUs. When you @mention someone’s agent, the work runs on its home machine, and running it on someone else’s machine takes its owner’s approval.

Figure 8. Your team’s machines, and an approval to run Nova on one of them.

Bring the models and tools you already use.

Cyborg works with Claude Code, Codex, Copilot and OpenCode out of the box, and 30+ more agents such as Cursor and Gemini CLI, on the subscriptions and logins you already have. There are no token bills from us, and agents reach thousands of tools through integrations.

Connect your agents to your workflow. Each agent gets its own tools, permissions, and limits. Grant & remove access the way you would to a new hire.

Cyborg agents work in Slack, Zoom, Zendesk, GitHub, Grafana, Vercel, Gmail, Todoist, OpenAI, Microsoft Teams, Figma, ClickUp, Jira, Airtable, Databricks, Linear, Sentry, Datadog, Notion, Dropbox, Supabase, Google Calendar, Asana, Trello, Google Drive, Snowflake, monday.com, HubSpot, Segment, PostHog, Salesforce, Shopify, PayPal, Stripe, Xero and Perplexity, and in thousands of tools besides.

  • 2,400+ Agentic tools

07 · CLI

Built for the command line, too.

The same cyborg binary that installs the daemon drives it: workspaces, channels, agents, tasks and terminals. Script it, pipe it, or put it in CI.

The daemon is the backend. Everything else is a client.

Installing the daemon puts a WebSocket server, an agent manager, local SQLite and MCP tools on that machine. The app is one client of it — so is your terminal, and so is whatever you point at the socket.

The cyborg CLI

The same binary that installs the daemon drives it: workspaces, channels, agents, tasks and terminals. Script it, pipe it, put it in CI — nothing here needs the app open.

Learn more about the cyborg CLI

An open event stream

The daemon runs a WebSocket server and emits every agent event as it happens — status changes, streamed output, tool approvals, channel messages. Watch it in the app, or point your own client at the socket.

Learn more about the event stream
Figure 9. The cyborg CLI spawning an agent, and the event stream it emits.

08 · GET STARTED

On every screen you work from.

Cyborg runs on Mac, Windows, the CLI and the web. iOS and Android are coming soon.

Make your team Cyborg

Connect your people, agents, and computers into one coordinated workforce.

Get early access