TermCanvas
Manage terminals, Git worktrees, and AI coding sessions on one infinite canvas.
- Source repo
- blueberrycongee/termcanvas
- Stars
- ★ 419
- Last updated
- 5mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 45/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificCodex · Claude Code
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Developers running several Claude Code or Codex sessions who want to see their states at a glance.
- Not a fit if
- Teams that want a browser-only tool without a desktop app
- Developers who do not organize projects with Git worktrees
- Source review
- 45/100 · Major gaps
What does this agent do, and when should you use it?
TermCanvas is an Electron desktop app for macOS, Linux, and Windows that places terminals on a pannable, zoomable canvas. It organizes work as projects → worktrees → terminals and detects a project's worktrees. The app supports Claude Code, Codex, Kimi, Gemini, and OpenCode, alongside shell, lazygit, and tmux terminals. Its interface includes session browsing, Git status and diff views, and usage tracking for Claude and Codex. Users can work in the app or register the termcanvas and hydra command-line tools; Hydra coordinates agent work through local workbench files. The README does not state the subscription or service costs associated with usage syncing.
After a user adds a Git project, TermCanvas detects its worktrees and lets them create terminals within them; newly created worktrees appear on the canvas. Users can pan, zoom, arrange, and star terminals, inspect agent status and session history, and view Git status, commit history, and diffs. After registering the CLI, they can run termcanvas project add, create shell or agent terminals with termcanvas terminal create, and inspect telemetry with termcanvas telemetry get. For coordinated work, hydra init-repo syncs instructions into CLAUDE.md and AGENTS.md, while hydra init, dispatch, and watch manage a workbench. Workbench state and result evidence are stored in the repository under .hydra/workbenches/.
- Developers running several Claude Code or Codex sessions who want to see their states at a glance.
- Engineers using Git worktrees to isolate branch tasks and monitor each worktree's status.
- Individual developers who want shell, lazygit, tmux, and AI coding terminals in one workspace.
- Technical leads assigning work to isolated terminal workers with Hydra and checking structured result files.
- Developers who need to replay past Claude or Codex sessions and inspect Git diffs in the same app.
How do you install or deploy this agent?
Download a desktop build from GitHub Releases. The README lists macOS, Linux, and Windows; Apple Silicon Mac users should choose a file with arm64 in its name. To build from source, the repository identifies pnpm-lock.yaml as its canonical lockfile:
git clone https://github.com/blueberrycongee/termcanvas.git
cd termcanvas
pnpm install
pnpm devAfter launching, go to Settings → General → Command line interface and click Register to install the termcanvas and hydra commands. Building from source requires Git and pnpm; the README does not specify a pnpm version.
How do you use this agent?
Add a project in the app, or register the CLI and run:
termcanvas project add ~/my-repo
termcanvas terminal create --worktree ~/my-repo --type claude --prompt "Audit the auth flow and fix the root cause"
termcanvas terminal status <id>To use a Hydra workbench, run hydra init-repo in the project, then initialize and dispatch work:
hydra init-repo
hydra init --intent "Add OAuth login" --repo .
hydra dispatch --workbench <id> --dispatch dev --role dev --intent "Implement OAuth login" --repo .
hydra watch --workbench <id> --repo .What are this agent's strengths and limitations?
- Uses a freeform canvas and project → worktree → terminal hierarchy to manage parallel terminals instead of relying on tabs or split panes.
- Supports Claude Code, Codex, Kimi, Gemini, and OpenCode, with agent status shown on terminal tiles.
- Hydra keeps workbench contracts in repository files and uses validated result.json as its machine gate.
- Combines session replay, Git status and diffs, and Claude and Codex usage views in the desktop app.
- Requires an Electron desktop app; the README documents no browser-based way to run it.
- Source builds use pnpm, and the app's workspace model centers on Git worktrees, which may not fit projects that do not use Git.
- Usage syncing requires sign-in, but the README does not specify service costs, account requirements, or sync details.
- CLI registration also installs Claude/Codex skills and lifecycle hooks; teams should review those integration behaviors and configuration.
How does this agent compare with similar options?
Compared with a workflow based only on terminal tabs or split panes, TermCanvas arranges terminals on a freeform canvas and organizes them by project and Git worktree. The README does not name a direct competing product.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| TermCanvas This agent | 45 · Major gaps | Desktop app | ★ 419 | 5mo ago | TypeScript | Codex · Claude Code |
| Skiff | 72 · Some gaps | Desktop appFree + model costs | ★ 20 | today | Rust | Codex · Claude Code |
| NeuroSquad Terminal Dashboard | 80 · Good | CLIFree + model costs | ★ 90 | 26d ago | TypeScript | Codex · Claude Code |
| Alethe — Multi-Agent Coding Workspace | 70 · Some gaps | Desktop appFree + model costs | ★ 1.1k | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The package centers on terminals, PTYs, workspace state, telemetry, and optional callbacks or webhooks; the README also says CLI registration writes Codex configuration. This shows broad capabilities and external effects, but the excerpts do not establish sufficient access controls, confirmation flows, data-flow disclosure, sensitive-data protections, dependency auditing, or rollback details, so these criteria receive 1. Author and repository metadata are traceable, but publisher identity is unverified and maintenance or update responsibility is not clearly established, so attribution receives 1.
The README, package configuration, and source describe a coherent product with a CLI, desktop app, and headless service, and show lockfile-based installation and failure logging. However, Hydra documentation uses differing workflow terminology, while the source sample gives limited evidence about service-boundary errors, dependency availability, and recovery behavior; each reliability criterion receives 1.
The README clearly targets cross-platform terminal and AI-coding-agent users, and lists terminal types, shortcuts, CLI scenarios, and installation guidance, supporting scores of 2 for audience and environment fit. Permission boundaries are not sufficiently described, while some skill triggers are broad or depend on user phrasing, so capability boundaries receive 1 and trigger precision 2.
The README is organized by feature and provides installation instructions, command references, workflow examples, and a user-guide link. Package metadata identifies the product and version, the release workflow requires a changelog entry, and the MIT license text is complete; information architecture, install notes, examples, and changelog receive 2, and license receives 3. The excerpts do not adequately document known limitations, naming compatibility policy, or maintenance responsibility, so those criteria receive 1.
Canvas-based terminal management, session resumption, Git panels, and Hydra orchestration provide clear product value. CLI commands and a structured result contract support practical use, but the workflows add complexity and the excerpts provide no cost or resource tradeoff analysis; output usability and marginal value receive 2, while cost-benefit receives 1.
The version, commands, and some implementation details are corroborated across the README, package metadata, and source, but most feature and safety claims remain README assertions, and the sampled tests do not cover those claims. Facts, product claims, and inference are not systematically distinguished, so all three verifiability criteria receive 1.
- The README describes optional callbacks, webhooks, telemetry, and Codex configuration writes during CLI registration; verify network destinations, written settings, and collected data before use.
- Hydra skills can launch or coordinate agent terminals and include reset, merge, and cleanup operations; inspect the workspace and result files before acting to avoid mishandling uncommitted work.
- The source excerpts do not establish dependency security, all error paths, or data-protection measures. This is a low-confidence static review, not an executed or independently tested assessment.
Maintain this project? Add the FARS badge to your README
[](https://followagents.com/en/agents/termcanvas)<a href="https://followagents.com/en/agents/termcanvas"><img src="https://followagents.com/badge/termcanvas.svg" alt="FARS review"></a>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
Fixed - Legacy Computer Use cleanup no longer removes user-managed Codex MCPs: Codex computer-use MCP entries are now removed only when they contain TermCanvas legacy signatures, preserving custom or third-party MCP servers that use the same generic name.View on GitHub ↗ -
Release notes excerpt
Changed - TermCanvas no longer ships its own Computer Use stack: removed the unfinished helper, MCP server, settings page, CLI commands, launch injection, and packaging path so Codex sessions use the bundled OpenAI Computer Use plugin instead. Fixed - Legacy Computer Use MCP entries are cleaned up on skill install: TermCanvas now removes stale Claude and Codex MCP registrations from older builds, including orphaned Codex TOML args / env lines that could keep pointing agents at the retired TermCanvas server.View on GitHub ↗ -
Release notes excerpt
Fixed - Unicode paths stay readable in Git-backed views: the file tree and Git diff now parse git ls-files output as NUL-separated records, so Chinese and other non-ASCII filenames render as real paths instead of escaped octal strings. - File tree refresh prunes stale empty directories: after tracked or ignored paths disappear, the file tree now removes empty ancestor folders that no longer exist in the refreshed path snapshot.View on GitHub ↗
FAQ
Is TermCanvas free to use?
Can I use it without Claude Code or Codex?
Does Hydra take control of agent sessions?
Does it require a network connection or sign-in?
Thanks — we use this to improve the page.