OpenCove Infinite Canvas Workspace
Put your AI agents, terminals, tasks, and notes on one infinite canvas for visible context.
- Source repo
- DeadWaveWave/opencove
- Stars
- ★ 1.6k
- Last updated
- 2d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 46/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude CodeChatGPT (Partial support)
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer who needs to run multiple Claude Code sessions simultaneously and compare their progress without losing track of each session's output.
- Not a fit if
- Teams that need stable, production-grade tooling rather than an Alpha-stage app
- macOS users who expect a signed, notarized app that installs without Gatekeeper workarounds
- Users who want the browser Web UI to work out of the box without password and LAN setup
- Source review
- 46/100 · Major gaps
What does this agent do, and when should you use it?
OpenCove is a spatial development workspace for developers who work with AI coding agents daily. Instead of burying work in tabs, sidebars, and long chat threads, OpenCove places AI agents, terminals, tasks, and notes on the same infinite 2D canvas, so the full state of your work stays visible. It is built with Electron + React + TypeScript, using @xyflow/react for fluid infinite canvas interactions, with xterm.js and node-pty powering real terminal PTYs. OpenCove is optimized for running multiple Claude Code or Codex sessions side by side, and it supports persistent workspaces that restore viewport, layout, terminal output, and agent state. It also offers space archives to snapshot and revisit previous workspace states, as well as global search and a control center for managing sessions. This is an Alpha-stage project offering prebuilt binaries for macOS, Windows, and Linux, plus an experimental Worker-hosted Web UI for browser access.
OpenCove provides an infinite canvas where you can place and arrange terminal nodes, notes, tasks, and AI agent sessions. It integrates terminal emulation (xterm.js + node-pty) to run CLI agents like Claude Code and Codex directly on the canvas. Users can run multiple agent sessions in parallel, view terminal output in place, and use global search to find content across the canvas and terminal output. The control center allows managing active sessions from one place. Workspaces are persistent, saving viewport and layout, and space archives snapshot workspace states for later reverting. The app also includes a headless server and Web UI (experimental) to access the canvas from a browser. Deployment uses the opencove CLI, with a headless server started via opencove worker start, and password-protected web UI.
- A developer who needs to run multiple Claude Code sessions simultaneously and compare their progress without losing track of each session's output.
- A developer working on a complex project who wants to keep task plans, personal notes, and terminal output in one visible area instead of switching between windows and tabs.
- A developer who pauses work on a feature to handle an urgent bug and later wants to restore the exact canvas layout, terminal output, and agent state to continue where they left off.
- A team that needs transparent monitoring of automated agent actions, ensuring every command and its effects are visible to observers.
- A developer switching between multiple projects who wants each project isolated in its own workspace to avoid context mixing.
- A user who wants to access their canvas from another device on the LAN via the web UI, e.g., to check progress from a tablet.
How do you install or deploy this agent?
Download the prebuilt binary for your OS from the GitHub Releases page. For macOS, since builds are unsigned, if Gatekeeper blocks, run: xattr -dr com.apple.quarantine /Applications/OpenCove.app. Alternatively, install the CLI via terminal: On macOS/Linux run: curl -fsSL https://github.com/DeadWaveWave/opencove/releases/latest/download/opencove-install.sh | sh, and on Windows PowerShell run: powershell -NoProfile -ExecutionPolicy Bypass -Command "Invoke-RestMethod https://github.com/DeadWaveWave/opencove/releases/latest/download/opencove-install.ps1 | Invoke-Expression". To build from source, ensure Node.js >= 22.12.0 and pnpm 9.6.0, then: git clone https://github.com/DeadWaveWave/opencove.git && cd opencove && pnpm install && pnpm dev.
How do you use this agent?
Launch the OpenCove desktop app and add terminal nodes, notes, and tasks on the infinite canvas. To use CLI agents, open a terminal node and run claude or codex. Arrange nodes to your liking. Use global search to find text across the canvas and terminal output. For web access, enable the experimental Web UI in settings, set a password, and start the local worker. For headless use, run: opencove worker start --hostname 0.0.0.0 --web-ui-password 'change-me' to serve the Web UI and allow password login.
What are this agent's strengths and limitations?
- Spatial organization of multiple agent sessions, terminals, and notes improves context visibility during parallel work.
- Modern tech stack (Electron, React, TypeScript, @xyflow/react) ensures smooth canvas interactions and performance.
- Persistent workspaces and space archives allow saving and restoring canvas state, including terminal output and agent state.
- Provides CLI installation and an experimental web UI, enabling access from a browser across devices.
- Alpha stage implies possible instability and incomplete features.
- macOS builds are unsigned, requiring manual Gatekeeper overrides.
- Web UI is experimental and defaults to loopback-only; LAN access requires additional configuration and password setup.
- Installing from source demands Node.js >= 22.12.0 and pnpm 9.6.0, which may be higher than some users' environments.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| OpenCove Infinite Canvas Workspace This agent | 46 · Major gaps | Desktop appFree | ★ 1.6k | 2d ago | TypeScript | Codex · Claude Code |
| Apache Maka | 82 · Good | Desktop appFree + model costs | ★ 5.8k | today | TypeScript | — |
| Kun Workspace | 49 · Major gaps | Desktop appFree + model costs | ★ 6.3k | 1d ago | TypeScript | ChatGPT · Codex · OpenAI API · Claude API |
| Superset | 52 · Major gaps | Desktop appFreemium | ★ 15k | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows a security policy (SECURITY.md) and dependency overrides (package.json), but no concrete implementation of least privilege, user confirmation, data flow transparency, or sensitive data handling. External effects (e.g., network, filesystem) are not clearly specified. Rollback and source attribution are only partially evident (e.g., version control).
The project has architecture audit tests and CI workflows, but the quality of failure messages is not assessed. Dependency availability is indicated via package.json and CI lock, but actual availability is not verified.
README clearly defines target users and scenarios, but capability boundaries and trigger precision are not detailed. Environment fit (macOS/Windows/Linux) is mentioned but not deeply.
Information architecture is clear (README, docs), install notes are detailed, but naming stability, examples/FAQ, known limitations, versioning/changelog, and maintenance responsibility are partially missing or insufficient.
Output usability (e.g., CLI, Web UI) is described, marginal value (vs traditional workflows) is articulated, but cost-benefit is not quantified.
Claims are partially traceable to code (e.g., architecture tests), but cross-source corroboration is insufficient, and fact-inference separation is unclear.
- Project is in Alpha stage, features may be unstable.
- macOS builds are unsigned, may trigger Gatekeeper warnings.
- Web UI defaults to loopback only, but enabling LAN access requires setting a password to avoid exposure.
- Dependency overrides and patches may introduce unknown risks.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
What's Changed * feat(ui): make the control center quick switches by @DeadWaveWave in https://github.com/DeadWaveWave/opencove/pull/457 * feat(ui): show the viewport Space and agent activity in the top bar by @DeadWaveWave in https://github.com/DeadWaveWave/opencove/pull/458 * feat(ui): show every project on the rail with its colour and activity by @DeadWaveWave in https://github.com/DeadWaveWave/opencove/pull/459 Full Changelog: https://github.com/DeadWaveWave/opencove/compare/v0.4.0-nightly.20261017.1...v0.4.0-nightly.20261018.1View on GitHub ↗ -
Release notes excerpt
What’s new OpenCove 0.4.0 introduces Pixel themes, a more flexible canvas, and independently installed plugins and themes, with improvements to terminal links and Agent recovery. - Choose Pixel or Pixel Light, customize window colors, and control pixel effects for the canvas, windows and Agent activity. ([#422](https://github.com/DeadWaveWave/opencove/pull/422)) - Collapse windows into compact desktop icons, restore them with smooth motion, and keep their collapsed state across restarts. ([#422](https://github.com/DeadWaveWave/opencove/pull/422)) …View on GitHub ↗