by nodo
attention-queue pricing, free plan and limits
Terminal app that queues every Claude Code session in tmux that needs you as one card at a time, with four checking modes: review, step, quiz, vibe.
- Best for
- Developer running several Claude Code sessions in tmux
- Not for
- macOS or Windows developer
Should you use attention-queue?
Pick it if you run several Claude Code sessions in tmux on Linux and keep losing track of which one is waiting on you. Skip it on macOS, on Windows or with other agents. The repository was created on 10 November 2026, so treat it as early software.
One keyboard queue that answers Claude Code permission prompts, questions and diff reviews across every tmux session, plus a quiz mode that checks you understood the change.
Lock-in is low: the code is MIT, its state sits in plain files under the user's config and state folders, and the hooks are plain JSON entries in settings.json.
Also for
- Developers running several Claude Code sessions in tmux on Linux
- Reviewers who want to read and comment on agent diffs in the terminal
- Engineers who want to check they understood an agent's change
Also not for
- Developer who works with Codex or OpenCode
Main trade-off
It is days old and works only with Claude Code on Linux inside tmux.
attention-queue pricing
No price, account or hosted service is listed in the README. The optional quiz and side-question features spend your own Claude Code usage.
Before you commit
- Quiz, judge and side questions run claude -p calls, which use your own Claude Code allowance (src/llm.rs)
What attention-queue does
attention-queue, or aq, is an open-source terminal app written in Rust that gathers what Claude Code sessions in tmux need from you, such as diffs, edit approvals, permission prompts and questions, into one queue of cards. It offers four per-project modes: review, step, quiz and vibe. It runs on Linux.
One queue across sessions
Gathers diffs, edit approvals, permission prompts and questions from every tmux session into one list, shown one card at a time.
Review with line comments
Lets you comment on a line or range of a finished turn's diff and send all comments back to the session as a single message, like a pull request review.
Step mode
Holds each edit until you press y to allow it or n to refuse it with a reason.
Quiz mode
Asks you to explain chosen parts of a diff in your own words, has a judge mark the answer and allows up to two retries with a leading question before the answer is shown.
Show all 8 features
Questions and option menus
Answers Claude's questions and option menus from a card, handing the answer back through the hook so nothing is typed into the session.
Session control
Shows a tab per session and lets you message it, ask where it is up to without interrupting, switch its model or permission mode, or start a new session.
First-run tour
Walks through review, step, quiz and sessions on made-up cards, with no path to any real session.
Shell popup
Opens a tmux popup shell in the project folder with Claude's last suggested command typed in, for sudo, logins or commands its permissions block.
Works with
- Claude Code hooks
- tmux
- git worktrees
- $EDITOR
Missing integrations
- Agents other than Claude Code
- macOS and Windows builds
Most users expect these and they are not currently supported.
Capabilities
- Claude code
- Tmux
- Code review
- Terminal UI
- Agent oversight
- Hooks
- Rust
Strengths and limits
Strengths
- The source is open and tested: MIT-licensed Rust, an end-to-end test suite run against a private tmux server according to the README, and a checksum file published beside the Linux release archive, with
cargo install --path .offered for anyone who would rather build it. - The README says that when aq is not running its hooks do nothing and Claude carries on as normal, so installing it does not make sessions depend on it.
- The hook notes are careful about provenance: docs/hooks.md states it was checked against Claude Code 2.1.319 on 2026-11-07 and separates facts from Anthropic's hooks reference from facts seen in real runs.
Limits
- It is very new: the repository was created two days before this page was written, and has 5 stars, no forks and, in a web search on 12 November 2026, no independent write-ups, so expect quick changes and little community support.
- It covers one setup only: Linux with tmux, one prebuilt x86_64 Linux file and Claude Code sessions, so macOS, Windows and other agents are not catered for.
- Step mode makes every edit wait for a person, which stalls a session that nobody is watching.
attention-queue alternatives
The nearest AI tools on Hokai for the same job.
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| attention-queue This page | Free | Yes | -- | 3.2 |
 Claude CodeAnthropic's agent for the terminal, IDEs, desktop, web and mobile that edits files, runs tests and opens pull requests. There is no free plan. | $20/mo | No | Web · iOS · Android · Mac · Windows | 4.6 |
 DevinDevin is Cognition's autonomous AI software engineer that plans, codes, tests, and ships pull requests, with a Free plan and several paid tiers. | $20/mo | Yes | Web · Mac · Windows | 4.6 |
 CursorCursor is an AI code editor built on VS Code, widely adopted across large enterprise engineering teams, with Agent Mode, Tab completion, and Cloud Agents. | $8/mo | Yes | Web · iOS · Mac · Windows | 4.4 |
attention-queue facts
- Works offline
- No
- Minimum skill level
- Intermediate
- Price band
- Free
- Maker founded
- 2026
Data and privacy
MIT
Questions about attention-queue
What does attention-queue cost?
Nothing is charged for aq: the README lists no paid plan, no account and no hosted service, and the code is MIT-licensed. The quiz, judge and side-question features run background claude -p calls, so they use your own Claude Code allowance. You also need tmux and Claude Code already installed.
Is attention-queue free to try without touching my real sessions?
Yes. The first run starts a tour on made-up cards, and the README says nothing in the tour reaches a real session. You can learn the keys there before adding the three hooks to your Claude Code settings.json. F2 runs the tour again and F1 lists every key.
Does attention-queue work on macOS or with Codex?
The README says aq runs on Linux inside tmux, and the only prebuilt file on the release is an x86_64 Linux archive. It is built on Claude Code hooks, so sessions in other agents are not covered. The README also offers a cargo build from source, but it makes no macOS claim.
Will installing aq break my existing Claude Code hooks?
Only if you replace them. The README says to merge the three aq hooks into your existing PreToolUse, PostToolUse and PermissionRequest arrays, because replacing the arrays stops your other hooks from running. Its paste-in install prompt tells Claude to back up settings.json and show the diff before saving.
What happens to an edit in step mode if I do not answer?
The edit is not applied. docs/hooks.md records a real run with a two-second deadline and a socket that never answered: Claude did not write the file and saw a message asking it to try the edit again shortly. The same page says that a hook giving no decision sends the call through the normal permission flow, and that silence never approves anything. Install aq if you run several Claude Code sessions in tmux on Linux and can watch step cards; otherwise pick another tool.
About attention-queue
attention-queue, shortened to aq, is a Rust terminal app for developers who run several Claude Code sessions inside tmux and lose track of which one is waiting on them. It connects through Claude Code hooks and turns whatever needs a human, such as a diff to review, an edit to allow, a permission prompt or a question, into a card shown one at a time. A card that arrives while you are on another waits its turn instead of taking the screen. It comes from nodo, a GitHub organisation with one public repository, which was created on 10 November 2026; the code is MIT-licensed, and v0.1.8, published on 11 November 2026, was the newest release when this page was written.
Read the full background
Each project has its own mode, switched with the M key. Review turns a finished turn's diff into a card where you comment on lines and send the comments back to the session as one message. Step holds every edit until you allow it or refuse it with a reason. Quiz picks parts of the diff and asks you to explain them, so the change is understood before it is accepted, and vibe checks nothing. Quiz questions, the judge that marks answers and side questions run as background claude -p calls (src/llm.rs, with a 180-second limit per call). Read on 12 November 2026, its Cargo.toml lists no HTTP client crate among its 13 dependencies. When no card is waiting, aq shows a tab per session, where you can message one, change its model, or start another in its own tmux window, optionally in a separate git worktree. It builds on Claude Code from Anthropic and supports no other agent.
Elsewhere the fit narrows, because aq works with Claude Code only, runs on Linux inside tmux and ships one prebuilt file, an x86_64 Linux archive. Anyone on macOS or Windows, or who prefers OpenCode or OpenAI Codex as the agent, is outside what the README covers; the Codex vs Claude Code guide helps with choosing the agent first. Its own README also says that someone who wants one app managing every session's branch is better served by the tools in its comparison table, Claude Squad, Conductor, Vibe Kanban and tuicr. Claude Squad (a terminal manager for several agents) and Vibe Kanban (a task board for coding agents) have no Hokai record yet. The wider AI coding assistants list shows what else Hokai covers for coding.
Sources and guides
Official sources
- attention-queue official site Vendor
- attention-queue README Docs
- Claude Code hook formats used by aq Docs
- Latest release Docs
- GitHub repository GitHub
Product facts on this page (pricing, plans, integrations, compliance) are read from nodo's own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.
Lists that include attention-queue
More AI Tools on Hokai
AtuinOpen-source shell history with encrypted sync across machines, plus Threads to search and resume coding-agent sessions (32.3k GitHub stars).
CodeAnt AIAI pull request reviewer with built-in SAST, secrets and quality checks across four Git hosts, used by 2,000+ organizations by the vendor's count.
GitarAI code review agent that diagnoses CI failures and pushes fixes to the pull request, free for open source and supporting 4 code hosts.
GitHub CopilotGitHub's AI coding assistant with agent mode, Copilot Fusion model routing and MCP support across 6 IDEs, plus a free plan with no credit card required.
LightrunAI SRE platform that adds live logs, snapshots and metrics to running production apps without redeployment, now with autonomous incident detection and remediation.
SourcegraphEnterprise-only code intelligence platform for cross-repository semantic search and AI-assisted code understanding across large, multi-repo codebases, after free and Pro self-serve plans were discontinued.