Nimbalyst Visual Coding Workspace
A desktop workspace for running Codex and Claude Code sessions in parallel while reviewing file changes visually.
- Source repo
- nimbalyst/nimbalyst
- Stars
- ★ 2k
- Last updated
- 4d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 72/100 · Some gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A developer progressing several Codex or Claude Code tasks at once can use parallel sessions, search/resume, and the Kanban board to keep them coordinated.
- Not a fit if
- Users who mainly want OpenCode or GitHub Copilot: both are documented as alpha and may be unstable
- Developers unwilling to install npm dependencies and build the Electron monorepo from source
- Source review
- 72/100 · Some gaps
What does this agent do, and when should you use it?
Nimbalyst is a local, open-source visual editor, session manager, and task manager for developers, product managers, designers, and builders. Its monorepo includes an Electron desktop app, a SwiftUI iOS app, cross-platform runtime services, collaboration-protocol types, an extension SDK, and built-in extensions. It supports Codex and Claude Code, with OpenCode and Copilot identified as alpha, and connects parallel sessions with files and task boards. Built-in editors cover Markdown, mockups, Mermaid, Excalidraw, CSV, data models, and Monaco-based code editing, with red/green visual review of coding-agent changes. Content and status can live in Markdown, ordinary files on disk, or Git; the collaboration sync server is a separate project.
Users create or open documents, edit them in a WYSIWYG Markdown editor, and use supported coding agents across their files. The app links sessions to files and files to sessions, groups files touched by a session, runs sessions in parallel, and provides search, resume, and Kanban management. When an agent suggests edits, the user reviews red/green WYSIWYG diffs, accepts or rejects changes, and can edit or annotate the material directly. Its task features let agents and humans add, move, and execute plans, bugs, features, and todos; developer features include Git-state management, AI git commit, an embedded Ghostty terminal, and worktrees. Custom editors plug in through the EditorHost contract; named built-in extensions include an Astro website editor, visual Git log, mindmap, slides, and a 3D object editor.
- A developer progressing several Codex or Claude Code tasks at once can use parallel sessions, search/resume, and the Kanban board to keep them coordinated.
- A technical lead reviewing agent-generated changes can inspect red/green diffs in Markdown, code, or visual-file workflows before accepting each edit.
- A product team working across specifications, mockups, diagrams, CSV files, and code can collaborate with coding agents in the corresponding built-in editors.
- An engineer using isolated Git worktrees can combine Git state, AI git commit, and the embedded Ghostty terminal in one workspace.
- Someone away from their desktop can use the mobile app to see active sessions, answer agent questions, review diffs, and queue follow-up tasks.
How do you install or deploy this agent?
Download the listed installer for macOS Apple Silicon, macOS Intel, Windows, or Linux. The documented minimum requirements are macOS 12+, Windows 10+, and Linux. To run from source, npm 7+ is required:
npm install
cd packages/electron && npm run devThe supplied documentation does not specify credential setup, authorization, or first-connection steps for Codex, Claude Code, OpenCode, or Copilot.
How do you use this agent?
Start by clicking New or pressing Cmd/Ctrl+N, then write or edit in the WYSIWYG Markdown editor. Switch to Agent Manager to run parallel sessions, link them with files, and manage progress through search or the Kanban board. Review each agent change through the red/green diff interface, accepting or rejecting it as appropriate; use the task view for plans, bugs, and todos. For source development, cd packages/electron && npm run dev starts the Electron app in development mode.
What are this agent's strengths and limitations?
- Combines sessions, files, tasks, and Git-oriented workstreams in one desktop application, including bidirectional session-file links.
- Provides visual collaboration surfaces for Markdown, mockups, Mermaid, Excalidraw, CSV, data models, and Monaco code.
- Lets users approve or reject coding-agent edits through red/green WYSIWYG diffs instead of relying only on terminal output.
- Uses a common
EditorHostcontract for built-in and custom editors, with several named extension types already included. - Offers listed desktop downloads for macOS, Windows, and Linux, alongside an iOS mobile-app workspace.
- OpenCode and Copilot are documented as alpha, so those integrations may be incomplete or less stable.
- The collaboration sync server is outside this repository; this codebase supplies the client and wire-format types in
packages/collab-protocol. - Source use requires npm 7+ plus dependency installation and Electron-monorepo development setup.
- The supplied documentation does not explain how to authorize or configure any supported coding-agent provider.
- Anonymous PostHog analytics are enabled by default for fresh installs, although users can opt out in Settings → Advanced → Analytics.
How does this agent compare with similar options?
Nimbalyst is positioned as a visual collaboration and session-management layer for Codex, Claude Code, OpenCode (alpha), and Copilot (alpha), rather than as a replacement for those coding agents. Compared with working directly in those agents' terminal sessions, it adds file linking, Kanban management, task tracking, visual editors, and diff approval.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Nimbalyst Visual Coding Workspace This agent | 72 · Some gaps | Desktop appFree + model costs | ★ 2k | 4d ago | TypeScript | Codex · Claude Code |
| Tessera | 70 · Some gaps | Desktop appFree + model costs | ★ 318 | 11d ago | TypeScript | Codex · Claude Code |
| dev-3.0 | 76 · Good | Desktop app | ★ 329 | today | TypeScript | Codex · Claude Code |
| Vicoa | 67 · Some gaps | Desktop appFree + model costs | ★ 531 | 1d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The permissions document describes per-project controls, workspace scoping, sensitive-path blocking, URL patterns, and several approval scopes. These support ordinary access control. Deductions: users can choose unrestricted access, and permanent approvals are stored in project settings; the MCP migration guide says it deletes the legacy config after migration but does not describe backup or recovery. Telemetry disclosures identify an anonymous ID, event-list location, and opt-out setting, but the event list and analytics guide are not included here.
Provider permission mappings, MCP migration behavior, product architecture, and installation requirements are documented; the CLI entrypoint also describes stable error handling. Deductions: CLAUDE.md says hooks and validation gates enforce many rules, but their implementations and run results are not in this packet; the MCP implementation document labels the work complete and reports 22 tests, which remain claims in the supplied material. Some documentation is explicitly outdated, and the MCP implementation excerpt does not show the full implementation.
The README identifies developers using multiple coding agents and explains parallel-session, visual-editing, task-tracking, and mobile scenarios, with platform requirements. Permissions and installation notes cover multiple modes and Linux distributions. Deductions: full capability boundaries are spread across area-specific documents; the Wiki plugin skills show trigger conditions, but the sample is incomplete and cannot establish coverage of all triggers.
The repository has package-level architecture guidance, build and installation steps, Linux troubleshooting, an MIT license, a version number, stable and alpha update channels, and documented limitations. Deductions: some documents are explicitly outdated or historical; maintenance responsibility is implied through team, contribution, and security contacts rather than clearly assigned with response commitments; the supplied material cannot establish the completeness of the changelog.
Product descriptions and onboarding steps explain concrete uses such as visual editing, diff review, parallel sessions, and task tracking; approval patterns can also be managed and removed. Deduction: combining desktop, multiple agents, MCP, extensions, and mobile features creates setup and learning costs, and the material provides too little evidence to judge whether those costs are worthwhile for each audience.
Specific README claims can be partially compared with the permissions document, CLI entrypoint, workflow files, and source excerpts; the permissions document gives code locations, and the identity document calls for line-level citations. Deductions: several important claims are documented but not corroborated by implementation or independent verification in this packet; the identity document says some of its evidence came from a working tree rather than HEAD, weakening confidence that it matches the specified revision.
- The “Allow everything” mode automatically accepts operations; review the selected permission mode and remembered approvals when trusting a project.
- The MCP migration guide says it deletes the old configuration after success; the supplied material does not explain how to restore that deleted file.
- Some architecture and implementation documents are explicitly outdated or historical; do not treat their status descriptions as a complete account of the specified revision.
Maintain this project? Add the FARS badge to your README
[](https://followagents.com/en/agents/nimbalyst)<a href="https://followagents.com/en/agents/nimbalyst"><img src="https://followagents.com/badge/nimbalyst.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 - New tracker items get local numbers (like NIM.75) again; items created since early October without one are numbered on the next launch. A tracker type can opt out with localNumbers: false - The pull request panel no longer reports the GitHub CLI as signed out when gh auth status is slow to answer. - After you start a new session to implement a plan, the planning session's plan card says "Deferred to new session" instead of "Denied" - A session that fails to load now says why and offers Try again, instead of showing "Session not found" …View on GitHub ↗ -
v0.81.7 Release v0.81.7
-
v0.80.6 Release v0.80.6
FAQ
Is Nimbalyst paid software, and what license applies?
Which coding agents are supported?
Is the collaboration backend included in this repository?
wss://sync.nimbalyst.com, but the Cloudflare Worker that powers that sync service is a separate project.Does the app collect source files or account information?
Thanks — we use this to improve the page.