“Worktree isolation alone paid for the license. Four agents, zero collisions.”
Pay once. Get the whole suite as it ships.
Pro and Team subscribers unlock every new Quorum product at no extra cost. No upsells, no per-seat add-ons for new tools.
BetaQuorum Terminal · macOS · available now
The terminal IDE forClaude Code, Codex & Geminimulti-agent workflows.
Not a code editor. A project-management framework for developers who ship with Claude Code, Codex CLI, and Gemini. Multi-agent terminals, worktree-safe sessions, and context that survives reloads.
14-day free trial·Apple Silicon + Intel·macOS 13+
Runs 100% locally·Zero telemetry·Your code never leaves your machine
session · live
claude code / src/main
usage · 1h
$2.41 / 18k tokens
branch
feat / usage-tracker
agents · 4
4 running / 0 idle
Workflow+
quorum
Grid · 4 running
aukcjonariusz-ai
Sessions · idle
mcp-agency-ads
Sessions · 1 wait
claudemain · usage
codextests
clauderefactor
+
Usage · live1h
Tokens18,247
Cost · est.$2.41
Burn · 1h$2.41/h
Reset2h 17m
Sessions4
usage-trackerclaude
flaky e2ecodex
release notesclaude
Mentioned in
Hacker NewsProduct HuntLobstersr/LocalLLaMAIndie Hackers
Honest numbers from a real, growing private beta
0+
Developers in private beta
since Jan 2026
0
Concurrent terminals supported
per project
0
Weekly releases shipped
since beta launch
4.8/5
Average tester rating
in-app feedback, n=312
312 builders shipping with Quorum today
No logo wall — we don’t have FAANG yet. Real handles, real workflows, verifiable on GitHub.
“I know what a feature costs BEFORE I ship it. Burn-rate clock = game changer.”
“AGENTS.md became our operations doc. We keep 4 worktrees per PR.”
“I open the laptop, open Quorum, type the goal, ship by lunch.”
“I went from 14 iTerm tabs to one window. I got my evenings back.”
“Eight agents on an M1 Air, 60fps, <1GB RAM. No idea how, but it works.”
“The per-terminal AI selector is mind-blowing. Claude for refactors, Codex for tests.”
“P90 + the depletion clock stopped 3 runaway agents for me last month.”
The problem
Vibecoding without a workspace is chaos.
You started in the terminal because that's where the AI actually lives. Three weeks in, you have 14 iTerm tabs, lost branches, untracked decisions, and no idea what your agents shipped last Tuesday.
Before
Stock terminal + a dozen tabs
Brancheslost across worktrees
Contextreset every reload
Costsinvisible until the invoice
Agentsoverwriting each other
Promptsretyped from memory
After
Quorum Terminal
Branchesone panel, one source of truth
ContextAGENTS.md, persistent
Costslive per-session, per-tool
Agentsisolated worktrees, by default
Promptssaved library, one click
Live workspace
A workspace, not a window.
Click between modes to see Quorum work the way you do — whether you're running four agents at once or staring down a 400-line diff.
Active sessions
usage-trackerclaude
flaky e2ecodex
diff: renderclaude
docs · onboardinggemini
+ new terminal
usage-tracker · 1claude
❯ claude "wire P90 to hit-limit blocks"► editing aiUsageLiveSnapshot.js► +24 lines · -8 lines✓ tests passing · 643/643❯
flaky e2e · 2codex
❯ codex "reduce retries to 2"► found 3 instances⏵ waiting for permission to edit⚠ /test/e2e/login.spec.js
diff: render · 3claude
❯ claude "explain hunks line-by-line"► reading git/diff.js► 412 lines · 7 hunks identified► drafting explanation
docs · 4gemini
❯ gemini "rewrite QUICKSTART"✓ done · 124 linesidle 12m
Branches
feat/usage-tracker●
main
fix/e2e-flakes
refactor/git-pane
+ new branch
142// Sliding 1h burn rate window-const burnRate = totalTokens / blockAgeMs;+const windowMs = 1 * 60 * 60 * 1000;+const sliced = records.filter(r => now - r.t < windowMs);+const burnRate = sliced.reduce(sum, 'tokens') / windowMs;147148// P90 filtered to hit-limit blocks only-const p90 = percentile(allBlocks, 90);+const hitLimit = allBlocks.filter(b => b.usage >= 0.95);+const p90 = hitLimit.length >= MIN_SAMPLE+ ? percentile(hitLimit, 90)+ : percentile(allBlocks, 90);
Changes3
+aiUsageLiveSnapshot.js
McodexUsageHistory.js
Mtest/aiUsage.test.js
Dlegacy/burnRate.js
feat(usage): P90 on hit-limit blocks + sliding burn rate
usage-tracker· feat/usage-tracker
$1.42
5.2k tokens · 23% ctx
❯ claude "wire P90 to hit-limit blocks · sliding 1h burn rate"► reading aiUsageLiveSnapshot.js · 312 lines► editing P90 baseline · 18 lines changed + hitLimitBlocks filter (≥95% ceiling) + 192h closed-block window - percentile over all blocks► editing burn rate · 14 lines changed + sliding 1h window - total/age extrapolation► running tests · jest --updateSnapshot✓ 643/643 passing · 1.4s · 0 regressions✓ ready to review · run git diff to inspect❯
4 active · 2 endedtoday
CL
usage-tracker
feat/usage · 4m 32s · 5.2k tok
$1.42
CX
flaky e2e
fix/e2e · waiting · 2.1k tok
$0.41
CL
diff: render hunks
main · running · 8.4k tok
$2.18
GM
QUICKSTART rewrite
docs · idle 12m · 3.7k tok
$0.08
CL
release notes 1.0
release · ended · 12.3k tok
$3.04
CX
migration script
db/migrate · ended · 6.8k tok
$0.92
What's inside
Built for the way AI actually writes code.
Every panel earned its place. Nothing decorative.
01 Multi-terminal
Up to 32 terminals per project
Tab, grid, or Sessions view. Layouts from 1×1 to 4×4 plus custom splits. Each terminal can run a different AI tool.
main
❯ claude
► editing…
claude
tests
❯ codex
⏵ waiting
codex
diff
❯ claude
► 7 hunks
claude
docs
❯ gemini
✓ done
gemini
02 Source control
Git that lives next to your prompts
Staged, unstaged, branches, worktrees, inline diffs. Commit without leaving the keyboard.
+aiUsageLiveSnapshot.js+24 -8
McodexUsageHistory.js+12 -3
Dlegacy/burnRate.js-87
03 Agents
Switch AI tools without switching apps
Claude Code, Codex CLI, Gemini CLI — one tool selector per terminal.
claude
codex
gemini
+ custom
04 Cost
Real-time usage and burn rate
P90 baseline, sliding 1h burn rate, predicted depletion clock. Pricing tables verified weekly.
Claude · usage62%
Codex · usage34%
Burn · 1h$2.41/h
05 Prompts
A library of saved prompts
Reuse the prompts that work. One-click paste, version history, per-project scoping.
refactor for testability
explain this hunk-by-hunk
write a migration script for…
06 Context
Context that survives the session
Your architectural decisions, agent instructions and project notes — alive across reloads.
AGENTS.md
STRUCTURE.json
PROJECT_NOTES.md
.cursorrules
CLAUDE.md
07 Safety
Worktree-safe multi-agent launches
Spin up two, four, or eight agents on the same codebase. Each gets its own detached worktree. No collisions.
08 Extensible
Plugin marketplace
Extend Quorum in JavaScript. Official marketplace plus local plugin support.
@quorum/jira
1.2k installs
@quorum/linear
2.8k installs
@quorum/figma-export
614 installs
+ build yours
Agent autonomy
You decide how much rope the agent gets.
Four discrete levels of autonomy — from read-only suggestions to full execution inside an isolated worktree. Switchable per terminal. The worktree is what makes level 4 safe.
L1no risk
Read-only
Agent reads the repo, describes what it would do, never touches files. Pure review mode.
Best for: code reviews · onboarding to a new repo
L2low risk
Suggest & show diff
Agent proposes changes as a patch. You accept or reject hunk-by-hunk. Nothing writes without your click.
Best for: refactors on production code · changes to core
L3mid risk
Approve each step
Agent runs autonomously, but every tool call (write, exec, git) needs your y/n. Shortcuts: A = approve-all for this session, Y/N per step.
Best for: feature work · your daily AI loop
L4safe in worktree
Full autonomy
Agent runs without asking — but ONLY inside an isolated worktree under ~/.frame/worktrees/. It physically cannot reach outside its checkout. Your main branch stays untouched.
Best for: parallel multi-agent runs · overnight tasks
Worktree isolation is what makes „Full autonomy” safe by default. Cursor, VS Code and Warp let an agent write straight into your branch — we don’t.
Also included
Twelve more tools you didn't know you needed.
Quorum keeps shipping. Every feature below ships in the current build.
Claude Code Skills
First-class Skills support — pick a skill from the picker, agent picks it up automatically.
Git activity heatmap
365-day commit heatmap per project. See what your AI shipped at a glance.
Inline conflict resolver
Side-by-side merge conflict UI with one-click "take theirs / ours / both" right next to the diff.
Element inspector
Annotate any HTML element from a live preview pane. Comments flow straight into your agent's prompt.
Auto-updater
Background updates with rollback. New version every Friday. No manual DMG re-downloads.
Theme presets
Tokyo Night, Dracula, Catppuccin, and a Quorum-native dark theme. Per-project override.
Terminal Chronicle
Timeline of every command, edit, and tool call per session. Searchable, scrubbable, exportable.
Workspace wizard
Guided setup picks Grid vs Sessions, terminal count, and AI tools — saves a launchable preset.
Session history backfill
Imports old Claude/Codex sessions from ~/.codex and ~/.claude logs. Resume anything you started last month.
Prompt history with timestamps
Every prompt logged, scoped per project, with one-click re-paste and search across sessions.
Keyboard-first
A keyboard-first command set. Switch terminal, toggle panels, focus tree, jump to Nth tab — without ever touching the mouse.
In-app CodeMirror editor
Quick edits without leaving Quorum. JS, TS, Python, Go, Rust, Markdown, JSON, YAML and more — full syntax + paths preserved.
Keyboard-first
Built by people who hate touching the mouse.
Every panel, view, and action lives one keystroke away. Hover any shortcut to highlight it.
Terminals
⌘⇧TNew terminal
⌘⇧NSplit terminal
⌘WClose terminal
⌘1–9Jump to terminal
Navigate
⌘[Previous terminal
⌘]Next terminal
⌘EFocus projects
⌘⇧EFocus file tree
Panels
⌘⇧HHistory panel
⌘⇧GSource control
⌘⇧SSaved prompts
⌘⇧PPlugins panel
AI & agents
⌘KStart AI tool
⌘⇧CCommit changes
⌘⇧AAgent waiting
⌘⏎Open file reference
How it works
Three steps. One workspace.
01
Open the folder
Drag a repo into Quorum, or open one from the launcher. AGENTS.md and STRUCTURE.json are detected automatically.
❯ open ~/projects/quorum► detected AGENTS.md · STRUCTURE.json► branch: main · worktree clean✓ workspace ready
02
Pick your agents
Choose Claude Code, Codex, or Gemini — for each terminal independently. Quorum sets the working directory and starts the tool.
term 1 · claude code · mainterm 2 · codex cli · teststerm 3 · gemini cli · docs✓ 3 agents started
03
Ship
Stage, diff, commit, push — without leaving the keyboard. Sessions remember themselves. Resume any time.
git: 3 staged · 0 unstagedcommit: "feat(usage): sliding 1h burn rate"✓ pushed to origin/feat/usage-tracker✓ session archived · resume anytime
Who uses Quorum
Built for four kinds of builder.
The Vibecoder
Solo · MVP shipping
You think in prompts, not in IDEs. You want one place to keep the chaos, with version control that doesn't make you switch contexts.
"I open the laptop, open Quorum, type a goal, and ship by lunch."
The Agent Orchestrator
Multi-agent · parallel work
You run four Claude sessions in parallel — one on docs, one on tests, one on the refactor, one watching the logs.
"Worktree isolation alone paid for the license."
The Indie Hacker
Cost-aware · ship daily
You watch every dollar. Live cost meters and burn-rate predictions mean no surprise invoices at the end of the month.
"I finally know what each feature costs before I ship it."
The Tech Lead
Team rollout · governance
You roll AI tools out across a team. You need audited prompts, shared context files, and consistent guardrails.
"AGENTS.md became our team operating doc."
Quorum vs the rest
Built for one workflow. Done well.
Editors that bolted a terminal on the side, terminals that forgot about version control, IDEs that forgot about agents. Quorum picks one job.
| Feature | Quorum Terminal | VS Code | Cursor | Warp |
|---|---|---|---|---|
| Native multi-agent terminalsUp to 32 per project, isolated worktrees | 32 | limited | tabs only | |
| AI tool selector per terminalClaude · Codex · Gemini · custom | single | |||
| Real-time usage & cost metersP90 baseline, burn rate, depletion clock | ||||
| Saved prompts libraryPer-project, versioned | extensions | |||
| Inline diff + commit panel | ||||
| Session resume across reloads | chat only | history | ||
| Built for the terminal first |
The Quorum Suite
One license. A growing family.
Quorum Terminal is the first product in a suite built around the same idea: ship faster, with AI, without losing your workspace. Subscribers get every tool we ship.
Public roadmapEvery product has a real release window. We don't ship vaporware.
Quorum TerminalLive today
Terminal IDE for AI coding
Multi-agent terminals, Git, prompts, usage tracking. The product you're looking at.
Get it nowQuorum BriefQ3 2026
Spec-first context for your agents
Generate AGENTS.md, STRUCTURE.json, and onboarding docs from a single prompt. Keep them in sync as the codebase moves.
Join the waitlistQuorum ReviewQ4 2026
Pull-request review for AI-shipped code
Detect drift between spec and diff. Flag prompts that produced regressions. Audit every agent commit.
Join the waitlistMore on the way · 2027 roadmap
Flow, Insights, Cloud — and the things you ask for.
Every Quorum Pro and Team subscription includes every new product in the suite at no extra cost. Vote on what we build next.
Flow · runbooks
Insights · analytics
Cloud · team sync
Pricing
Free to try. Honest when it isn't.
14 days free, no card. Subscribers get every Quorum product as it ships.
Most popular
Pro
$15.20/mo per dev
Billed annually · $182.40/yr.
Save $45/yearEvery Quorum product — Terminal today, Brief / Review / Flow as we ship them.
- Up to 32 terminals per project
- Unlimited concurrent agents
- Worktree-safe launches
- Real-time usage + burn rate
- Plugin marketplace + saved prompt library
No card requiredCancel anytime
Team Coming soon
—
For teams of 3+. Pricing announced at launch.
- Everything in Pro · for the team
- Shared prompt & context libraries
- SSO + audit logs (Q3 2026)
- Priority support · 1-day SLA
- Roadmap input + early features
Invoice / VATVolume pricing 10+
Compare all features across plans
| Feature | Propopular | Team coming soon |
|---|---|---|
| Core terminal | ||
| Concurrent terminals | 32 | 32 |
| Split layouts | up to 4×4 | up to 4×4 |
| Sessions view | ✓ | ✓ |
| In-app CodeMirror editor | ✓ | ✓ |
| AI & agents | ||
| Claude / Codex / Gemini support | ✓ | ✓ |
| Worktree-safe multi-agent | ✓ | ✓ |
| Saved prompts library | unlimited | unlimited |
| AGENTS.md autoload | ✓ | ✓ |
| Cost & usage | ||
| Per-session cost tracker | ✓ | ✓ |
| Burn-rate P90 + depletion clock | ✓ | ✓ |
| Org-wide spend dashboard | ✕ | ✓ |
| CSV export & invoicing | ✓ | ✓ |
| Team & admin | ||
| Shared prompt library | ✕ | ✓ |
| Centralized billing | ✕ | ✓ |
| SSO (SAML, OIDC) | ✕ | Q3'26 |
| Audit logs & SCIM | ✕ | Q3'26 |
| Custom DPA | ✕ | ✓ |
| Support | ||
| Email response | 48h | 1 business day |
| Roadmap input | ✕ | ✓ |
All plans include local-first execution, Apple-notarized builds, and GDPR-aligned EU data processing.
Pricing doesn’t fit? Team of 10+? Compliance audit?
Let’s talk. Volume pricing from 10 seats, custom DPA, on-premises licensing, NDA code audit — all real, all with a human, not a chatbot.
Apple-notarized · signed & stapled DMG
Local-first · zero telemetry by default
GDPR-aligned · EU data processing
Built in Warsaw · EU company, VAT-invoiced
After download
From DMG to your first agent in 90 seconds.
Three commands. No config, no project setup, no toolchain dance.
1
Install & open a repo
Open the DMG, drag to Applications, point Quorum at any local repo. AGENTS.md is detected automatically.
open ~/projects/my-app2
Pick an AI tool
Toolbar selector lists every CLI you have installed — Claude, Codex, Gemini, or your own. Each terminal can run a different one.
claude code3
Ship
Stage, diff, commit, push — all from the keyboard. Sessions resume across reloads, so you never lose state.
⌘ ⇧ G → commitWords from beta
Builders who already moved in.
"Quorum is the first dev tool in two years that made me close VS Code on day one. The session resume is uncanny."
"Running 6 Claude sessions on the same repo without worktrees was insane. Quorum just… handles it."
"The cost meter is the only thing that stopped me from accidentally spending $400 on Opus in an afternoon."
"Saved prompts changed how my whole team writes. We have a shared library now and our PRs read consistent."
"It looks like a Linear product and feels like a Vim plugin. Whoever built this gets it."
"AGENTS.md autoload alone earned a spot in my dotfiles workflow. Best 19 bucks I spend monthly."
From the founder
SKSebastian · founder
Hi. I built Quorum because I was tired of my workspace.
For 18 months I shipped my own products with Claude Code, Codex, and Gemini. I had eight terminal tabs open at all times, lost branches I couldn't remember creating, surprise invoices from runaway agents, and prompts I retyped from memory because I couldn't find the good ones.
Quorum is the tool I wanted on day one. It is opinionated about the terminal, generous about your existing tools, and pathologically honest about cost.
Shipping notes
Built in the open. Shipped every week.
No marketing roadmaps, no "coming soon" vapor. Just what we shipped, what we're shipping next, and when.
Changelog
All releases →Steadier session tiles and a self-healing update check
The context-usage bar on session tiles no longer disappears while cost attribution catches up, and a slow signature check at launch no longer switches auto-update off for the whole session.
Updates you can see — version, progress and notes on every update surface
The update experience got a full pass: every surface now says which version you are getting (with download progress in MB and release notes in Settings), a new sidebar card and an animated version pill announce updates clearly, and locally built (unsigned) copies of the app no longer break auto-update with a cryptic signature error.
Updates show up faster
The app now polls for a new version every 5 minutes instead of hourly, so a shipped update is offered to you much sooner.
Roadmap
Plugin marketplace v1
Discover, install and update JavaScript plugins from the in-app marketplace. Signed-by-author manifests.
SSO + audit logs (Team)
SAML / OIDC, SCIM provisioning, exportable audit logs. Required for enterprise rollouts.
Reproducible builds
Bit-for-bit reproducible release artifacts. Verifiable by anyone, anywhere.
Windows & Linux beta
Native builds for Windows 11 (signed) and Linux (AppImage + flatpak). Same codebase, same shortcuts.
System requirements
Will it run on my machine?
Quorum is a lean desktop app. We test on a 2020 M1 Air (8 GB) with 8 concurrent agents and stay under 1 GB RAM.
- Operating system
- macOS 13 Ventura or newer (Sonoma, Sequoia)
- Architecture
- Apple Silicon (M1+) & Intel x86-64 · native build per architecture
- Memory
- 4 GB minimum · 8 GB recommended for 8+ agents
- Disk space
- Compact install — exact size shown on the download card
- Node.js
- Optional — only if you use Node-based CLI tools
- Network
- Required only by the AI providers you connect
- Windows / Linux
- Q4 2026 · join waitlist
Security & privacy
We take your code as seriously as you do.
Quorum runs on your machine. No code, prompts, or session data ever touches our servers. The whole stack is signed, scoped, and auditable.
Local-first by default
Quorum is a desktop app. Your code, terminals, prompts and provider keys live only on your machine — never proxied through us.
- No code uploads, ever
- No prompt or session telemetry
- Provider tokens stored in OS keychain
Signed & notarized builds
Every release is Apple Developer-ID signed, notarized and stapled. Hardened runtime is on. SHA-256 checksums published with each download.
- Apple Notarization · macOS 13+
- Hardened runtime + library validation
- Reproducible builds (on the roadmap)
GDPR & EU data
Quorum is operated from Warsaw. Account data (email, license, billing) lives in the EU region. DPA available on request. Right to be forgotten honored in 14 days.
- EU-region account & billing data
- DPA + sub-processor list on request
- Privacy-first analytics, opt-in only
Closed source, fully accountable
Quorum is proprietary, closed-source software. We don’t need a license to prove the app doesn’t phone home — the architecture does. Open the in-app Network Inspector and watch every outbound call in real time. Enterprise customers can request a code audit under NDA.
- Proprietary code, signed binaries
- In-app Network Inspector — verify yourself
- NDA code audit on request (Team plan)
Verify it yourself
Open the Network Inspector. See what we see (nothing).
Every outbound HTTP request the app makes — update pings, license check, your own AI provider traffic — lands in this panel. With timestamp, destination, and payload size. Closed-source app, audit-grade transparency.
- Live stream, no buffering
- Export to JSON / HAR for audit
- Block any domain at the OS level
Providers you talk to directly — we never proxy a single byte
AnthropicClaude CodeOpenAICodex CLIGoogleGemini CLIBring your own CLIany terminal tool
Your API keys · your account · your invoice. We don’t even know which models you use.
All systems operational · quorumkairos.ai/status
FAQ
The questions that came up most.
Why a terminal IDE and not a regular editor?
When you build with AI, the agent writes the code — you spend most of your time in the terminal. Quorum is for that workflow. It's not trying to replace VS Code; it's replacing the eight terminal tabs you have open next to it.
Which AI tools does it support?
Claude Code, Codex CLI, Gemini CLI today, plus any terminal-based AI tool. Each terminal can run a different tool, and the AI selector lets you switch from the toolbar.
Is Quorum open source?
No — Quorum is proprietary, closed-source software. We chose to keep the codebase closed so a small team can ship a polished product without governance overhead. Instead of a license guarantee, we give you architectural ones: signed builds with published SHA-256, the in-app Network Inspector that shows every outbound call in real time, opt-in code audit under NDA for Team customers, and a transparent CVE process. Your code, prompts and sessions never touch our servers — that’s enforced by architecture, not by license.
Windows and Linux?
macOS first. The codebase is cross-platform-ready (node-pty and xterm.js work everywhere), and Windows and Linux are on the roadmap for Q4 2026.
How does worktree isolation work?
When you launch a multi-agent workspace, Quorum can create detached Git worktrees under ~/.frame/worktrees//... so each agent terminal writes to its own checkout. Branches are never auto-created.
How accurate is the usage tracker?
Pricing tables for Anthropic and OpenAI are verified weekly. The tracker reads local provider JSONL logs and shows API-equivalent estimates — not provider billing truth. Sliding 1h burn rate, P90 baseline filtered to hit-limit blocks.
Do you train on my code?
No. Quorum runs locally. We never receive your code, your prompts, or your usage data. The AI tools you run inside Quorum follow their own provider policies.
What does the subscription cover?
Every product in the Quorum suite — Terminal, and every Quorum tool we ship after — for as long as you subscribe. Pro and Team include all future products at no extra cost.
Download
Get Quorum on your machine.
macOS available today — native builds for Apple Silicon and Intel. Windows and Linux Q4 2026.