cate - Moses Yebei

The Problem

Running multiple coding agents (Claude Code, Codex, etc.) in parallel produces a mess of terminal tabs with no visibility into which agent is blocked waiting for input, which is done, and which is still working. Coordinating agents across git branches on one machine typically means manual worktree juggling and constant context switching.

What This Does

Cate is an Electron-based infinite canvas IDE where terminals, editors, browsers, and viewer panels live as spatial objects. Agent-aware terminals detect turn start/end and permission prompts via a protocol (piRpcClient.ts), so each panel shows running/waiting/finished state and pings you when an agent needs input. One click creates a git worktree with its own colored canvas territory, keeping parallel branches visually separate.

The cateAgent module (src/cateAgent/main/, src/cateAgent/renderer/) is the agent orchestration layer: session management, auth, model pickers, chat UI, and subagent orchestration. A CLI (src/cli/cate.ts) lets agents drive the browser panel, read other terminals, and manage panels from inside a terminal.

How It Is Wired

Execution starts in index.html → the React renderer. The Electron main process (src/main/) owns windows, drag logic (dragLogic.ts), browser panels (browser/agentBrowser.ts), and extensions (extensions/ExtensionManager.ts).

The agent control path: renderer UI → src/cateAgent/renderer/cateAgentSend.ts → IPC (src/cateAgent/main/ipcCoding.ts) → codingManager.ts, which routes through piRpcClient.ts to talk to the external agent process over RPC. authManager.ts handles provider credentials; codingDir.ts and seedWorktreeChat.ts own worktree creation. The cateAgent/extensions/ directory holds four installable modes—cate-orchestrator, cate-plan-mode, cate-canvas-mode, cate-ask-user—each with its own index.ts and tests, installed via installOrchestrator.ts and siblings.

The widest blast radius sits in codingManager.ts (session lifecycle, all agent state) and cateAgentStore.ts (renderer state). Both are heavily tested—399 test files total, including concurrent/race tests (authManager.concurrent.test.ts, dockWindowFlush.races.test.ts). The store-to-manager-to-RPC chain is the path any new agent feature must touch.

How To Use It

# Install (recommended, from README)
brew install --cask cate
# Or download a prebuilt release for macOS/Windows/Linux

# Build from source (per CONTRIBUTING.md)
git clone https://github.com/moses-y/cate
cd cate
npm install
npm run dev

No configuration files required—open a folder and it becomes a workspace. Press Cmd+K for commands, panels, and files. Shortcuts are rebindable in Settings.

Real-World Use

Run five Claude Code agents on five branches. Each gets a worktree with its own canvas territory; Cate surfaces which agent is waiting on a permission prompt. When one blocks, you answer it in place without leaving the canvas. An agent can call cate from its terminal to open a file or drive a browser panel for verification.

Code Health & Issues

Deep analysis has not run for this repo yet; the following is from structure only.

Low - Fork adds no README attribution to upstream 0-AI-UG/cate; verify license provenance before commercial use.
Med - No measured coverage data; 399 test files exist but pass/fail state requires running CI.
Low - research/browser-agent-architecture.md and plan.md suggest active design work; docs may drift from code.

CI (ci.yml), release automation (release.yml), MIT license, and a lockfile are all present.

The Bottom Line

A well-tested, genuinely useful tool for parallel agent workflows—the agent-aware terminal protocol and worktree-per-canvas-territory model solve a real coordination problem. The fork is fresh (0 stars vs. 2106 upstream), so treat it as a snapshot of upstream rather than an independent project. Use it if you run multiple coding agents daily; fork upstream directly if you plan to extend it.

What the analyser found

Deployment readiness

6/7

Container image
CI pipeline
Lockfile committed
Test suite
README
License
No committed secrets

Composition

1,066 files

TypeScript770
TSX209
Markdown18
JSON13
YAML4
Shell3

ReactTailwind

Module dependencies

The 10 most depended-upon modules of 989, from static import analysis. Red outline marks a module in an import cycle.

graph LR n0_src_shared_types["types"] n1_nderer_stores_appStore_index["index"] n2_src_main_logger["logger"] n3__renderer_stores_canvasStore["canvasStore"] n4_enderer_stores_settingsStore["settingsStore"] n5_src_shared_ipc_channels["ipc-channels"] n6_src_shared_runtimeLocator["runtimeLocator"] n7_src_main_runtime_types["types"] n8_src_renderer_lib_logger["logger"] n9_rc_renderer_stores_dockStore["dockStore"] n7_src_main_runtime_types --> n0_src_shared_types n7_src_main_runtime_types --> n6_src_shared_runtimeLocator n1_nderer_stores_appStore_index --> n0_src_shared_types n1_nderer_stores_appStore_index --> n6_src_shared_runtimeLocator n3__renderer_stores_canvasStore --> n0_src_shared_types n3__renderer_stores_canvasStore --> n8_src_renderer_lib_logger n9_rc_renderer_stores_dockStore --> n0_src_shared_types n4_enderer_stores_settingsStore --> n8_src_renderer_lib_logger n4_enderer_stores_settingsStore --> n0_src_shared_types classDef cyc stroke:#ef4444,stroke-width:2px; class n0_src_shared_types,n1_nderer_stores_appStore_index,n3__renderer_stores_canvasStore cyc;

Dependencies

16 of 50 declared dependencies are a major version behind. Resolved 50 of 50 against the registry.

Package	Declared	Latest	Behind
`@vitejs/plugin-react`	^4.3.0	6.0.5	2 major
`electron`	^41.2.0	43.4.0	2 major
`eslint-plugin-react-hooks`	^5.2.0	7.1.1	2 major
`globals`	^15.15.0	17.11.0	2 major
`typescript`	^5.6.0	7.0.2	2 major
`@xterm/xterm`	^5.5.0	6.0.0	1 major
`chokidar`	^4.0.0	5.0.0	1 major
`react`	^18.3.0	19.2.8	1 major

What would stop me shipping this

Ranked by severity × confidence × production reach. Reach is the honest discriminator across a collection that is mostly other people's code: the same finding matters more in something that ships.

high1

medium6

low2

Pin third-party GitHub Actions to a commit SHAhigh2 occurrences

.github/workflows

docker/setup-qemu-action@v4, peter-evans/create-pull-request@v8

A tag can be moved, so the action running with your token and secrets is whatever its owner last pushed; this is how tj-actions/changed-files leaked secrets from thousands of repos.

Fix: Replace each @vN with the 40-character commit SHA, keep # vN as a comment, and let Dependabot bump the SHAs.

Declare least-privilege permissions for GITHUB_TOKENmedium

.github/workflows/ci.yml

1 workflow(s) declare no permissions

With no declaration the token inherits the repository default, so any injected step can push commits or mint releases from inside your own CI.

Fix: Add permissions: contents: read at the top of the workflow and widen per job only where needed.

Enable Dependabot or Renovatemedium

5 manifest(s), no update bot configured

Without a bot a published advisory sits unpatched until someone audits by hand, which across 1,322 repositories means never.

Fix: Commit .github/dependabot.yml covering the repo ecosystems plus github-actions.

Gate pull requests on a dependency vulnerability scanmedium

.github/workflows

no dependency scan in CI

This is the one gate that would catch a known-vulnerable package before it reaches a build, and no repository in the sample had it.

Fix: Add dependency-review-action on pull_request, or osv-scanner on push and a schedule.

Move large binaries to Git LFS or out of the repositorymedium

assets/demo-canvas.gif

1 blobs over 5MB: demo-canvas.gif 8.9MB

One repository here carries twenty blobs over 5MB including a 10.9MB spreadsheet, so every clone and every CI checkout pays for data nobody diffs.

Fix: Track those extensions with LFS, or move datasets to object storage and fetch them in a setup step.

Set persist-credentials: false on checkoutmedium

.github/workflows/ci.yml

checkout keeps the token, then dependencies are installed

The token stays in .git/config for every later step, so a malicious postinstall script reads a pushable credential without one ever being passed to it.

Fix: Add with: persist-credentials: false, and pass an explicit token only to the step that pushes.

Review the install lifecycle script and disable scripts in CImedium

package.json

postinstall

Install scripts are the execution vector for every npm compromise from event-stream onward, and a postinstall that fetches a binary makes the build depend on a URL nobody reviews.

Fix: Move the work into an explicit build step, or set ignore-scripts in CI and run it by name.

Set timeout-minutes on the workflow jobslow2 occurrences

.github/workflows/release.yml

2 workflow(s) declare no job timeout

A wedged step runs to the six-hour platform default, which on a two-hourly schedule means three runs overlap behind it.

Fix: Add timeout-minutes with a realistic bound to each job.

Add the repository convention files this project lackslow3 occurrences

missing .editorconfig, .gitattributes, a formatter config

Without them one contributor's editor writes tabs into a Python file, a shell script commits with CRLF and fails in the container, and a notebook diff is unreviewable.

Fix: Add .editorconfig, .gitattributes with text=auto eol=lf, and a formatter config.

Checked deterministically against the repository tree and a bounded set of its files: committed credentials, unpinned actions and base images, missing lockfiles and update bots, workflows that discard failures, published advisories against the declared dependencies, runtime configuration, licensing and notebook reproducibility. No language model is involved in this section.