Novadeck

A desktop terminal that runs Claude Code, Codex and Antigravity side by side, with one Inbox for when they need you.

Stars
★ 55
Last updated
today
License
GPL-3.0
Primary language
TypeScript

At a glance

How it runs
Desktop appCLI
Works with
Portable with changesCodex · Claude Code
Cost
Free, no paid service needed
Setup effort
Medium · a few setup steps
You'll need
Claude Code CLICodex CLIAntigravityShell / CLINetwork accessLocal filesystem
Typical use
Developers juggling several Claude Code and Codex sessions who want all permission prompts in one place instead of alt-tabbing between windows.
Not a fit if
  • Developers who want a single terminal and no shared inbox
  • Environments that must stay fully offline and cannot run external CLIs
  • Teams needing a purely browser-native app rather than a desktop app
Website
novadeck.dev ↗ set on GitHub
Maintenance
Commits in the last 30 dayslast commit 2026-11-20latest release v1.9.53 (2026-11-20)+2.6 ★/day over 7 days

What does this agent do, and when should you use it?

Novadeck is a cross-platform desktop terminal app for running several coding agents — Claude Code, Codex and Antigravity — at the same time. Every agent's permission prompt, plan and question lands in a single Inbox, and most are answered with one click such as Allow, always allow or Deny. Agents message each other by handle and spin up their own teammates, and Claude Code, Codex and Antigravity can share one team. The UI offers Focus, Grid and Canvas views for one, a few or the whole crew, and pages, screenshots and diffs an agent shows you open in a pane next to its terminal. After a restart terminals reopen where they were and agents pick up their sessions. It runs the CLIs you already have with your own config and accounts — nothing is wrapped or rebuilt — and it is free, open source under GPL-3.0-or-later, with no account and no telemetry.

Novadeck ships as an Electron desktop app whose embedded terminal executes the Claude Code, Codex and Antigravity CLIs already installed on your machine, without wrapping or rebuilding them. While agents run, their permission prompts, plans and questions are routed into a shared Inbox where you answer in place — e.g. approving a Claude Code agent's request to run pnpm test:e2e, after which the tests run and pass. Agents message one another by handle and launch their own teammates across Claude Code, Codex and Antigravity, with a Messages pane showing a teammate's brief, question and the lead's answer. Views switch between Focus (one big terminal), Grid (tiled) and Canvas (all ten at once), and the companion pane opens the page, screenshot or diff an agent surfaces beside its terminal. On restart a splash counts the terminals it attaches, reports Deck online, and cuts back to the deck as it was.

  1. Developers juggling several Claude Code and Codex sessions who want all permission prompts in one place instead of alt-tabbing between windows.
  2. Teams where a Claude Code lead delegates to Codex and Antigravity teammates and needs a pane to read each teammate's conversation.
  3. Engineers running a dozen terminals who want to see the whole board at once in Canvas.
  4. Frontend developers who want an agent's page previews, screenshots and diffs displayed right next to the terminal for quick review.
  5. Solo developers who shut down often and expect terminal layout and agent sessions to come back exactly as left.

Key setup facts

Pulled from the install and usage notes below. The project README has the current values.

Install / run
  • curl -fsSL https://novadeck.dev/install.sh | sh
Install notes cover
Windows · macOS · Linux

How do you install or deploy this agent?

On macOS and x86_64 Linux, install with the one-line script:

curl -fsSL https://novadeck.dev/install.sh | sh

For Windows and portable builds, use the download page:

https://novadeck.dev/download

Prerequisite: the CLIs you intend to use (Claude Code, Codex, Antigravity) must already be installed and signed in on your machine, since Novadeck runs your own config and accounts.

How do you use this agent?

Open Novadeck after installing; it attaches the agent terminals you already have. When an agent needs you, the ask opens in the Inbox, for example:

Inbox → Checkout tests (Claude Code) asks to run: pnpm test:e2e
[Allow] [Always allow] [Deny]

Use Focus, Grid and Canvas to switch between one or many terminals, and the Messages pane to follow agent-to-teammate conversations. After a restart, terminals attach one by one, the splash reports Deck online, and sessions resume where they left off. To try it before installing, open:

https://try.novadeck.dev

What are this agent's strengths and limitations?

Pros
  • One Inbox collects every agent's permission prompts, plans and questions, mostly answerable with a click.
  • Agents message each other by handle and start teammates, with mixed Claude Code, Codex and Antigravity teams.
  • Runs the CLIs you already have with your own config and accounts, so nothing is wrapped or rebuilt.
  • Focus, Grid and Canvas cover the range from a single terminal to a full crew on one canvas.
  • Restart restores terminal layout and lets agents pick up their sessions.
  • Free, GPL-3.0-or-later, no account and no telemetry.
Limitations
  • Core functionality depends on the external Claude Code, Codex and Antigravity CLIs, which you must install and sign into separately.
  • The one-line installer covers only macOS and x86_64 Linux; Windows and portable builds go through a separate download page, and details are not in the repository.
  • No model/provider abstraction layer is documented, so adding other coding agents would require your own integration work.
  • The install script pipes a remote script into sh, so you must trust the remote; verification details are not provided.

How does this agent compare with similar options?

The README names no specific competitor; it only contrasts its side-by-side terminals plus shared Inbox with running agents without a unified inbox.

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
Novadeck This agent 55 · Major gaps Desktop appFree ★ 55 today TypeScript Codex · Claude Code
Rovai AI 77 · Good — ★ 276 1d ago Rust Codex · Claude Code
Poracode 67 · Some gaps Desktop appFree + model costs ★ 123 today TypeScript ChatGPT · Claude Code
Oh My OpenAgent 64 · Some gaps Agent plugin / skillFree + model costs ★ 71k today TypeScript Claude Code

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Major gaps
55/ 100 5-point scale 2.8 / 5
This review covers the source as of Nov 12, 2026. The project has shipped 4 stable release(s) since, which have not been re-reviewed. The score and verdict describe the reviewed version only. Recent releases ↓
Trust 17/29
Reliability 8/14
Adaptability 10/18
Convention 9/18
Effectiveness 7/13
Verifiability 4/8
Why each dimension lost points
Trust17 / 29 · 2.9/5

SECURITY.md documents per-channel sender validation, field allowlists, path checks and pkexec-instead-of-sudo for the desktop bridge (appearance, notices, uninstall, companion webviews), which is solid least-privilege and confirmation design; however dependency security rests only on a Dependabot automerge workflow with no lockfile audit or vulnerability-scan evidence, so dependency_security scores 1; source attribution is self-asserted (author mzpkdev, unverified publisher), scoring 1.

Reliability8 / 14 · 2.9/5

AGENTS.md commits to in-place workspace schema migrations and preserved settings/window state across self-updates, and the code comments are consistent with SECURITY.md, so self_consistency scores 2; dependency availability is only pinned via node 26.x, pnpm 11.22.0 and external toolchains with no fallback notes, scoring 1; failure messages are concrete and actionable in the whisper reviewed tests and SECURITY.md, scoring 2.

Adaptability10 / 18 · 2.8/5

README clearly targets developers running Claude Code, Codex and Antigravity side by side; AGENTS.md's Harness Parity section defines capability boundaries; trigger precision is weak because AGENTS.md is contributor guidance rather than invocation rules, scoring 1; environment fit covers macOS/Windows/Linux plus portable builds, scoring 2.

Convention9 / 18 · 2.5/5

README is well structured with a full docs index, scoring 2; install is a curl-pipe script plus a download page with no checksum or signature verification, scoring 1; naming stability is addressed (formerIds), scoring 2; examples and FAQ are thin, scoring 1; known limitations appear only scattered in SECURITY.md (e.g. no Touch ID passkeys on macOS), scoring 1; LICENSE is the full GPL-3.0 text matching package.json, scoring 3; version is still 0.0.0 with no CHANGELOG, scoring 1; maintenance responsibility names an unnamed 'owner' and the publisher is unverified, scoring 1.

Effectiveness7 / 13 · 2.7/5

Clear product positioning with screenshots and a demo link gives output usability 2; marginal value for parallel multi-agent workflows scores 2; cost-benefit lacks resource, performance or comparison data and install depends on an external script, scoring 1.

Verifiability4 / 8 · 2.5/5

Most README claims (no telemetry, no account) are self-asserted without verifiable privacy-doc content or code evidence, so claim_traceability scores 1; cross-source corroboration is limited to partial overlap between SECURITY.md and code samples, scoring 1; SECURITY.md explicitly requires separating confirmed behavior from suspicion, so fact_inference_separation scores 2.

Risks and how to mitigate them
  • Installation pipes a remote script into a shell with no checksum or signature verification; users should review the script first.
  • Publisher identity is unverified and the author field differs from the org name, leaving update and maintenance responsibility unclear.
  • Dependency security relies only on Dependabot automerge; no lockfile audit or vulnerability-scan evidence is present.
  • Version remains 0.0.0 with no CHANGELOG, making release stability hard to judge.
  • This is a static source review with no build or execution performed; confidence is low.
Evidence confidence: Low Reviewed Nov 12, 2026 Reviewed revision 3a28555891ca New commits since this review; the score may not cover them
See the full review method →
Maintain this project? Add the FARS badge to your README

The badge shows the current review score and verdict and links back to this page. It refreshes about once an hour after a re-review.

FARS

[![FARS review](https://followagents.com/badge/novadeck.svg)](https://followagents.com/en/agents/novadeck)
<a href="https://followagents.com/en/agents/novadeck"><img src="https://followagents.com/badge/novadeck.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.

  1. v1.10.5 Novadeck v1.10.5 Pre-release
    Release notes excerpt
    Nothing in this build changes what you see in the app. <!-- novadeck:notes {"version":"1.10.5","kind":"release","highlights":[],"changes":[],"fixes":0} --> <details><summary>All commits</summary> [1.10.5](https://github.com/novadeck-org/novadeck/compare/v1.10.4...v1.10.5) (2026-11-20) Bug Fixes * harnesses: a terminal's close leaves Claude Code's shared background service running ([#1143](https://github.com/novadeck-org/novadeck/issues/1143)) ([6cd859d](https://github.com/novadeck-org/novadeck/commit/6cd859d5b76c07fc5e0eaf9dd2529f0431f306e2)) </details>
    View on GitHub ↗
  2. v1.9.53 Novadeck v1.9.53
  3. v1.8.2 Novadeck v1.8.2

FAQ

Does Novadeck call models or charge for API usage itself?
No. It runs the Claude Code, Codex and Antigravity CLIs already on your machine under your own config and accounts, so usage is billed by those providers; Novadeck itself is free and account-free.
What happens after I click Allow in the Inbox?
The ask arrives with context, e.g. Checkout tests, a Claude Code agent, requests pnpm test:e2e; choosing Allow or always allow lets the agent continue, and its tests then run and pass.
Do agent sessions survive a restart?
Yes. On startup Novadeck attaches each terminal, reports Deck online, and cuts back to the deck as it was, with the Claude Code session where it left off.
Can I use it on Windows?
The README lists macOS, Windows and Linux, but the one-line installer covers only macOS and x86_64 Linux; Windows and portable builds are obtained from novadeck.dev/download.
Can agents from different vendors work together?
Yes. Claude Code, Codex and Antigravity can share a team, messaging by handle and starting their own teammates.
Is Novadeck free?
Yes. Novadeck is free to use and does not need a paid service.
Is Novadeck open source?
Yes. The source code is on GitHub (novadeck-org/novadeck) under the GPL-3.0 licence. GPL-3.0 is a copyleft licence: if you distribute a modified version, it must stay under the same terms.
Is Novadeck safe to use?
FollowAgents' FARS review (Nov 12, 2026) scored Novadeck 17/29 on trust and safety. None of the eight safety controls the review checks for was missing outright. The repository has changed since that review.
Is Novadeck still maintained?
Yes. The latest commit on GitHub is from 2026-11-20, within the last 30 days. The latest release is v1.9.53 (2026-11-20).
Who makes Novadeck?
Novadeck is published on GitHub by the account novadeck-org. FollowAgents has not verified which person or company is behind that account.
What is Novadeck's official website?
The project's GitHub repository lists https://novadeck.dev/?utm_source=github as its website. That link is set by the repository owner; the source code itself is on GitHub (novadeck-org/novadeck).
Was this page helpful?
View on GitHub ↗ Install ↓

Host ecosystems

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related concepts