One hub for every coding agent

Manage multiple AI coding agentsSync tasks to your phone

Run dozens of major AI coding agents side by side, connect all your devices, and continue the same task from your phone.

Every agent CLI—Dozens of major coding agents supported. Switch to another supported agent and the task keeps its full context. Agent harness support included.

Full mobile access—Continue the conversation, switch agents, track progress, and open generated work directly.

Open your workspace
Loading animation…

Multi-agent support

Switch AI coding agents inside the same task.

Fifteen agents are available today. Expand the directory to browse all 48.

Codex

codex
Available

Claude Code

claude
Available

OpenCode

opencode
Available

Gemini CLI

gemini
Available

Google Antigravity

agy
Available

Cursor Agent CLI

agent
Coming soon

GitHub Copilot CLI

copilot
Available

Kiro CLI

kiro-cli
Available

Kimi Code

kimi
Available

Qwen Code

qwen
Available

CodeBuddy Code

codebuddy
Available

DeepSeek Harness

dsh
Available

TraeCode CLI

traecli
Coming soon

Pi Agent Harness

pi
Available

Qoder CLI

qoder
Coming soon
Browse all 48 agents
Agents 16–48

Integrated agents are marked Available; all others remain Coming soon.

16

MiMo Code

mimo
Coming soon
17

Trae Agent

trae-cli
Available
18

CodeArts Agent CLI

codearts
Coming soon
19

MiniMax Code

mcode
Coming soon
20

Comate CLI

comatecli
Coming soon
21

CodeFuse Agent

cfuse
Available
22

MetaCode

metacode
Coming soon
23

Cline CLI

cline
Available
24

OpenHands CLI

openhands
Coming soon
25

Open Interpreter

interpreter
Coming soon
26

Goose

goose
Coming soon
27

Aider

aider
Coming soon
28

Kilo Code CLI

kilo
Coming soon
29

Freebuff

freebuff
Coming soon
30

Droid CLI

droid
Coming soon
31

Devin CLI

devin
Coming soon
32

JetBrains Junie CLI

junie
Coming soon
33

Mistral Vibe CLI

vibe
Coming soon
34

IBM Bob Shell

bob
Coming soon
35

Google Jules Tools

jules
Coming soon
36

Amp CLI

amp
Coming soon
37

GitLab Duo CLI

duo
Coming soon
38

Rovo Dev CLI

acli rovodev
Coming soon
39

Grok Build

grok
Coming soon
40

Auggie CLI

auggie
Coming soon
41

Warp Oz CLI

oz
Coming soon
42

mini-SWE-agent

mini
Coming soon
43

ForgeCode

forge
Coming soon
44

gptme

gptme
Coming soon
45

SWE-agent

sweagent
Coming soon
46

VT Code

vtcode
Coming soon
47

Every Code

coder
Coming soon
48

Qodo Command

qodo
Coming soon

Product capabilities

See how OpenTandem works.

Agent and model switching

Choose the right agent for the next run.

Switch between supported agents and their available models from the task composer. The task keeps its conversation and outputs.

  • Switch agents and models in one place
  • Models reported by the connected Runner
  • Live quota visible beside each agent

Mobile and artifacts

Move to your phone and keep working on the same task.

Mobile gives you the same task controls as desktop: continue the conversation, switch agents, track execution, and open generated images, documents, and HTML pages.

  • Keep the same task and context
  • Switch agents and models
  • Track progress and open outputs

Multi-machine orchestration

Know which machine is ready before you assign the task.

OpenTandem shows connected Runners and their live status. Send work to the machine that already has the right repository, tools, and environment.

  • Runner availability in real time
  • Project environment stays on the selected machine
  • One workspace tracks every execution

Task management

Manage tasks, conversations, and outputs in one place.

Organize work by project, keep conversations saved and archived, pin important tasks for follow-up, and open generated work from any device.

  • Organize tasks by project
  • Save, archive, and pin conversations
  • Preview task outputs across devices

Team collaboration

Share tasks, execution progress, and finished work.

Projects and tasks live inside a workspace with team access. Other members can continue the same execution without rebuilding its context in another conversation.

  • Workspace and member management
  • Shared projects and task progress
  • Conversations and finished work in one place

About OpenTandem

The workspace around your AI coding agents.

OpenTandem does not replace your AI coding agents. It keeps their tasks, machines, context, and outputs organized in one place.

01

One task context

Keep the task, conversation, execution progress, and finished work in one workspace.

02

Execution stays on your machines

A Runner executes on the machine that already has the right repository, tools, and environment.

03

Built for individuals and teams

Use every core capability on your own, then add shared projects, tasks, progress, and outputs for a team.

Pricing

Now open to everyone.

Public pricing has not been announced. Sign in with your email to open the complete OpenTandem workspace today.

Current access

OpenTandem Preview

Open access

Use the agent access already configured on each connected machine, without moving your repositories into another environment.

Open workspace

Included in the preview

  • Multiple connected machines
  • 15 agents available today
  • Agent and model switching
  • Live quota visibility
  • Full mobile access
  • Individual and team workspaces

CHANGELOG

Latest October 2026 Updates

Discover the latest monthly new features in OpenTandem, from interface and sign-in options to conversations, diagrams, plugins, Agents, messaging, and reply annotations.

  • OpenTandem CLI 0.3.2 Adds Individual Runner Control

    OpenTandem CLI 0.3.2 lets you select a Runner by ID on an already connected computer and start, stop, or restart it individually while other Runners keep running. This applies only to CLI-managed connections, while desktop-app-managed connections must still be managed in the desktop app.

  • Pair your computer from the command line with a QR code

    OpenTandem's new command line tool connects a macOS, Windows, or Linux computer without opening the desktop client: install it, run the pairing command, scan the QR code shown in your terminal with your phone, and sign in to approve the connection. The connected computer runs in the background, can start automatically at login, and accepts coding tasks sent from your phone or the web, provided supported coding agents are installed and the machine stays online.

  • Runner 0.91.0 continues native agent sessions across follow-up messages

    Runner 0.91.0 adds native session continuation for Cline, CodeFuse, DeepSeek Harness (version 0.1.6 or later required), and OpenClaw, so follow-up messages in the same conversation can continue the agent's own session instead of rebuilding the conversation context each round. Each agent still needs to be installed and configured on the connected computer, and this release does not add manual context compression, cross-agent session sharing, or guaranteed recovery of missing native sessions.

  • Search conversation titles across your workspace

    Sidebar search now finds unarchived conversation titles on computers in your current workspace that you have permission to access, rather than only on the selected computer. Select a result to open it; available on desktop and mobile web.

  • Navigate drives and enter folder paths in the directory picker

    On connected Windows computers, the directory picker can show drive roots under This PC so you can navigate to another drive when using Runner 0.85.7 or later. Directory pickers on Mac and Windows also include a Folder field for entering a directory path manually, including on mobile web; folders are selected on the connected computer, not from phone storage.

  • Google sign-in

    Google sign-in is available only in supported web browsers that can reach Google: web login supports Continue with Google, and the public website may offer Google One Tap. Existing accounts use their verified email and new users can register after accepting the agreement; embedded desktop and mobile clients do not show Google sign-in, desktop clients use the system browser, and email-code sign-in remains available when Google cannot be reached.

  • Understand your project

    On the desktop and mobile web new-conversation page, a starter card can fill the composer with a request to read the current directory and briefly explain the project. Editing is optional, and you can manually send the request when ready.

  • OpenClaw integration

    On a connected Mac with Runner 0.83.1 or later, you can choose an installed and configured compatible OpenClaw that supports agent exec. It uses OpenClaw’s configured default model, with results shown when the task completes and no live tool timeline; one-click installation is not provided, Windows is unsupported, and Linux has not been verified.

  • Independent font size preferences

    Settings now lets you separately set interface, conversation content, and code font sizes from 12 to 16 px, or restore the defaults. These preferences are saved only in the current browser and device, without account synchronization, and work on desktop and mobile web.

  • Return to bottom

    When you are reading earlier messages away from the bottom, a button takes you back to the latest messages and resumes following new content.

  • Hermes Agent integration

    On a connected computer running Runner 0.82.0 or later, select an already installed and configured Hermes Agent. It uses the model configured in Hermes, and its output appears after the run completes rather than as a live tool timeline.

  • Five-level numbered lists

    The message composer supports five numbered-list styles: 1., 1), a., i., and (1), with Tab and Shift+Tab to indent or outdent. Press Enter to continue numbering; the feature edits plain-text prefixes rather than rich text.

  • Resizable desktop conversations

    Drag either edge of the conversation column to adjust its width, and the messages and composer adapt together. Your width preference is saved to your account; resizing by dragging is available on desktop only.

  • Agent update management

    The machine list shows available updates based on each Agent’s installation source. Supported installed Agents can be updated from the UI with progress shown during the update; the machine must be online and use a compatible Runner and supported installation source.

  • Queued message tray above the composer

    While a task is running, additional messages appear above the composer. Selecting Insert sends the chosen message and any preceding queued messages to the current task in order, while your own messages that have not started can be returned to the draft for editing or deleted.

  • Getting started guide

    The guide lets you select an Agent already installed on a connected machine and choose a project directory, or skip project creation to start a conversation. You can return to the guide from the sidebar.

  • Switch Markdown tables to card view

    Switch Markdown tables in conversation replies and document previews between table and card views, and freeze or unfreeze columns through a chosen column on desktop and mobile web.

  • Save and share conversation images

    Turn a supported conversation reply, Markdown document preview, or table into a PNG image, preview and save it, or send it through the system share menu in browsers that support file sharing.

  • Answer native questions in runs

    When supported Codex, Claude Code, or OpenCode runs ask a native question, select an offered answer or enter your own where supported in the conversation card, then submit it to continue the run.

  • Install agents from the Agent list

    Choose a supported agent in the Agent list and start installing it on your connected computer, with a window showing installation output and status.

  • Reference another conversation with @

    Type @ in the composer to search for and reference an existing conversation, then ask your current agent to send a message there to hand off work.

  • Quote and comment on selected text

    Select text inside an AI reply, add an optional comment, and send both with your next message to give precise feedback without retyping.

  • Mermaid Diagram Views

    View Mermaid content in AI replies as diagrams, and switch between diagram, source, and zoom views.

  • Chinese and English Interface

    Choose Chinese or English in account language settings and save your preferred application interface language.

  • Conversation Handoff

    Ask the Agent to create a new conversation with a task summary and reference transcript for continued work.

  • Mobile QR Sign-In

    Scan the mobile usage QR code displayed on a signed-in desktop page to sign in on your phone.

View all features

Quick start

Start using OpenTandem.

Connect a Runner, choose an agent and model, then start your task.

Get started

FAQ

Questions, answered.

The practical details about agents, execution, mobile access, and team collaboration.

Which AI coding agents does OpenTandem support?

Fifteen coding agents are available now: Codex, Claude Code, OpenCode, Gemini CLI, Google Antigravity, GitHub Copilot CLI, Kiro CLI, Kimi Code, Qwen Code, CodeBuddy Code, DeepSeek Harness, Pi Agent Harness, Trae Agent, CodeFuse Agent, and Cline CLI. The expandable directory covers all 48 tracked agents.

Does the work run on my own machine?

Yes. A connected OpenTandem Runner starts the selected coding agent in the project environment on your machine.

Can I switch agents or models during a task?

Yes. Pick a different supported agent or model for the next run—the task keeps its conversation and outputs.

Can I use OpenTandem from a phone?

Yes. On mobile, you can continue the same task, switch agents and models, track execution, and open generated images, documents, and HTML pages.