A
One board, every thread
Wrangler joins each Linear ticket to its GitHub PR, its CI runs, and its rollout state — one reconciled row per feature instead of four half-truths in four tools.
Eyes on every agent
Every session,
tamed.
Wrangler is the control tower for engineers running a fleet of coding agents. Tickets, PRs, CI, and Slack — reconciled into one board you glance at a few times a day. See what needs you, act in one click, get back to real work.
●3 agents working ●1 waiting on you ● 0 surprises
01 / The stampede
You're running five agents and eleven tabs.
Parallel coding agents multiplied your output — and your surface area. Now every feature is a ticket in Linear, a PR on GitHub, a CI run somewhere, and a thread in Slack. None of them agree, and one of your agents has been quietly waiting on you for forty minutes.
Each glance costs a context switch. Each context switch costs the thing you actually sat down to do.
unreconciled · 7 signals · 4 tools
LIN-482 · In ProgressPR #1291 · review requestedCI ✗ failed — flaky again?#eng-alerts: “deploying?”agent-3: awaiting approvalPR #1287 merged… which ticket?agent-1: still working (12m)02 / The control tower
Glance. Act. Get back to work.
One row per feature — ticket, PR, CI, and rollout reconciled into a single status. One rail for your agents — so you always know who's blocked on you and who's happily grinding.
Wrangler · feature board● live
Checkout retry logicLIN-482PR #1291CI ✓staged 40%
Billing webhooks v2LIN-475PR #1287CI ◌—on track
Session replay opt-outLIN-491PR #1294CI ✗—
Org-level API keysLIN-460PR #1281CI ✓verified ✓on track
one-click, zero switches:ApproveRe-run CIReply in Slack
03 / The daily loop
From ticket to green board,
without the tab circus.
Wrangler doesn't replace your agents or your subscriptions — it sits above them. Bring the tools you already pay for; it handles the orchestration, the isolation, and the interruptions.
01
Connect your tools
GitHub via your existing gh login or a token; Linear or Jira for tickets; Slack, Coralogix, and LaunchDarkly if you use them. Tokens go straight into the OS keychain — they never touch the webview and only ever travel to each service's own host.
gh auth ✓ · linear ✓ · slack ✓
02
Start a task
Paste a ticket URL, a Slack thread, or just describe the work. Wrangler assembles the context, creates an isolated git worktree for the branch, and opens a live agent session — Claude Code, Codex, Cursor, Gemini, OpenCode, and 30 more.
worktree: lin-421-export-v2
03
Get pinged only when it matters
Agents grind in the background — detached sessions even survive quitting the app. A chime and a native notification fire the moment one finishes or needs input, and the "Waiting for me" inbox ranks what's blocked on you, most urgent first.
● agent done · needs review
04
Review, fix, ship
Read the diff the agent produced, attach line comments, and send them back as one batch. CI failed? One click dispatches a fix to the session that owns the PR. Approve, finish the feature, and the board goes green.
review → fix → approve → ✓
04 / In the field
The real app, on a real day
Not mockups — screenshots straight from Wrangler's own walkthrough suite. This is the tool we use to build the tool.
The board
Everything in flight, one glance
One row per feature: the Linear/Jira ticket, its PRs and CI, the rollout stage, and a single honest health status — reconciled automatically from all your tools. The red strip on top counts exactly what needs you.
Start a task
A ticket, a thread, or a sentence
Point it at a Linear or Jira ticket, paste the Slack conversation that kicked things off, pick the repo, pick the agent and model — Wrangler assembles the full context, spins up an isolated git worktree, and starts the session.
Sessions
Your whole fleet, live
Every agent session — even ones you started in a plain terminal — with its branch, model, message count, linked ticket and PRs, and a live status: running, waiting on you, or done. Open any of them as a real terminal, right in the app.
Needs attention
Triage, not archaeology
Failing CI, review comments, blocked work — ranked by urgency, scoped to what's actually yours. Each item carries a one-click action, so clearing the queue takes a minute, not a context switch.
Review changes
Read the diff, comment in batch
Inspect exactly what the agent changed, attach comments to lines across files — they collect in a pending panel — and send them all back as one structured instruction. The updated diff re-opens when the agent is done, so you verify — you don't hope.
One-click fix
CI failed? Dispatch the fix
The ⚡Fix button hands the failing check to the agent session that owns the PR — in its own worktree, with the autonomy level disclosed up front and push off by default so you always review before anything leaves your machine.
And the tricks nobody else has
Side chat
A second opinion, pre-briefed
Open a parallel session on the same feature — it reads the briefing, the other sessions, and the main transcript-so-far, so you can plan or ask without interrupting the run.
Memory
Clip it before it scrolls away
Select text in any session terminal and save it — the command that worked, the three files that matter. Kept 5 days, grouped by feature, gone before it becomes clutter.
Knowledge Center
Teach it once, every agent knows
Your project conventions — where logs live, which flags to check — plus facts the app learns on its own, merged into every session, side chat, task, and worktree CLAUDE.md.
⌘K
One palette over everything
Features, local tasks, agent sessions, and commands in one ranked list — by title, ticket id, repo, or branch. Enter jumps straight there.
05 / Field equipment
Built for the engineer who is now a fleet commander.
B
Who's waiting on you
Every agent session tracked in parallel. Working, idle, or blocked-on-you — with the exact thing it needs (an approval, an answer, a decision) surfaced up front.
C
Act without switching
Approve the PR, re-run the flaky CI job, reply to the Slack thread — right from the board. No tab safari, no "where was I?" tax when you get back.
D
Linear optional
Full picture with GitHub alone — PRs, CI, deploys. Connect Linear and Slack when you want tickets and threads folded in too. No all-or-nothing setup.
06 / Nothing ships unseen
A stable release is a story with no plot holes.
Wrangler walks every feature through the full arc — and keeps its eye on the joins, because that's where surprises hide.
1
Planning
ticket scoped
2
Coding
agent on it
3
Review
PR approved
4
CI
green, verified
5
Deploy
shipped
6
Rollout
staged %
7
Verified
case closed
caught before rollout
PR #1287 claims it closes LIN-291 — but the diff touches billing/ while the ticket scopes onboarding. Wrangler flags the mis-joined pair the moment it appears, not three days later in a post-mortem titled “how did that ship?”
07 / Saddle up
Bring your own agents.
Pay only for the tower.
Wrangler adds no AI bill — it drives the subscriptions you already have. And while we're in beta, every tier is unlocked for free.
Free
The single-player core, no strings.
$0
forever
- ✓The board, inbox & lifecycle tracking
- ✓Live agent sessions in-app (all 35+ agents)
- ✓Worktree-per-task isolation
- ✓GitHub (PRs, CI, one-click actions)
- ✓Up to 2 projects
Most popular
Pro
For the parallel-agent power user.
$10
per month · $96/yr · or $149 once (lifetime)
- ✓Everything in Free
- ✓Unlimited projects
- ✓Linear / Jira, Slack, Coralogix, LaunchDarkly
- ✓AI fix & review dispatch (⚡ one-click jobs)
- ✓Production monitoring on shipped features
Team
Shared visibility for agent-fleet teams.
$15
per seat / month · 3 seats min
- ✓Everything in Pro, for every seat
- ✓Shared team conventions & knowledge
- ✓Seat management with email invites
- ✓Fleet visibility (rolling out)
- ✓Priority support
beta = everything free · payments run in Stripe test mode · no card required to download
08 / Who we are
Built by an engineer
drowning in agents.
Wrangler started as a personal fix: one engineer running five parallel coding agents across Linear, GitHub, CI, and Slack — and losing an hour a day just working out what needed him.
So we built the control tower we wanted: a board that answers “what needs me right now?” in one glance, and gets out of the way the rest of the time. Wrangler is dogfooded daily — every feature on this page was shipped by agents wrangled inside Wrangler itself.
— the Wrangler crew · shipped from inside the app
Local-first, always
Your code, your transcripts, and your board data live on your machine. Tokens go in the OS keychain and only ever travel to each service's own host through an allowlisted proxy. No telemetry, no analytics, no phoning home.
Bring your own subscriptions
You already pay for Claude Max, Cursor, or Codex. Wrangler drives those — it never resells AI or adds a per-token bill. The thing you pay us for is the tower, not the planes.
Work, not sessions
Session managers show you terminals. Wrangler reconciles the work — the ticket, the PR, the CI run, the rollout — because "is this feature actually done?" is the question that matters at the end of the day.
09 / Get Wrangler
Put a watchful eye
on your fleet.
Free to download, free through the beta. Two minutes from DMG to a board that knows what needs you.
macOS 13+ · Apple Silicon & Intel · auto-updates built in
Beta builds are still being notarized — if macOS warns on first open, right-click the app → Open.
1
Download the DMG
Grab the latest build from GitHub Releases — Apple Silicon and Intel.
2
Drag to Applications & open
First launch runs a 2-minute onboarding: connect GitHub (your gh login is auto-detected), pick Linear or Jira, confirm who you are.
3
Start your first task
Paste a ticket, pick a repo, hit Start session. Updates arrive automatically from then on — the app offers “Restart & update” when a new version ships.
You'll want: git · one coding-agent CLI (Claude Code recommended) · gh for zero-config GitHub · tmux if you want sessions that survive quitting the app.
Questions we always get
?Is my code or my tokens leaving my machine?
No. Wrangler is local-first: tokens live in the macOS Keychain, requests go only to each service's own host through an allowlisted proxy, and there's no telemetry. The only outbound calls are the ones you connected.
?Which coding agents does it work with?
Claude Code first-class (live terminals, resume, headless jobs), plus Codex, Cursor, Gemini, OpenCode, Grok, and 30+ more. Headless one-click fix/review works with any agent that has an autonomous CLI mode.
?Do I need a new AI subscription?
No — that's the point. Wrangler drives the agent subscriptions you already pay for. It adds zero AI cost of its own.
?What about Windows and Linux?
macOS ships first (it's where we dogfood daily). The app is built on Tauri, and Windows/Linux builds are on the roadmap once the macOS beta settles.