Your agents speak up. Loud.
When a session stops to ask a question, the task moves into a red-bordered, pulsing column. You see it from across the desk — no Slack ping, no terminal switching.
Coming later this week
For builders running agents in parallel
One backlog. Many agents. One window. Run Claude agents like your own dev team: scope tasks, orchestrate the work, watch them build.
The problem you're already living with
You spin up agents in five terminals, lose track of two, and the third has been waiting for a yes/no for an hour. Sightdeck AI fixes that — without taking the IDE out of your hands.
.agents/sightdeck/sightdeck.xml — survives sessions, diffable in git.The promise
Built for devs who run agents all day — and need to stay in control without staring at terminals.
Sightdeck AI shows live for every Claude session: working, testing, waiting for you. Which tool is running, how long, in which project. No tab-switching, no log-tailing, no "is it still running?"
When a session is waiting for your reply, it jumps into a pulsing red column — visible even when Sightdeck AI runs in the background. You never lose an agent to an unanswered question again.
Tasks and concepts live as XML on your machine or even in your Repo. No telemetry, no vendor lock-in. You can uninstall Sightdeck AI today — your tasks stay.
Sightdeck AI assigns a Claude model to each task based on complexity. No expensive model for a text tweak, no cheap one for a bigger overhaul. Save token costs without thinking about it.
Drop a rough requirement in and Sightdeck AI breaks it into individual tasks with effort estimates and model suggestions. Structuring happens while you're writing, not in a meeting.
Select ten tasks, one click. Sightdeck AI checks which tasks touch the same files and decides parallel or serial. You start a whole run, not individual sessions.
Paste a Jira ticket URL — Sightdeck AI pulls in the key and acceptance criteria. Every task runs on its own Git branch, and the moment it hits done your build, Slack or status board hears about it. Your existing stack stays exactly as it is. Sightdeck AI just connects the dots.
Step 1 — turn thoughts into tasks
Throw everything on your mind into the quick-add box — a paragraph, three sentences, a Slack rant pasted in. Type it or dictate it with an AI tool — Sightdeck AI reads the whole thing, splits it into independent tasks, refines each one with a clean title, summary and estimate, and drops the batch into your backlog. Multi-select, hit Start, watch one agent per task work in parallel.
"Add VAT to the invoice PDF, translate the confirmation mail to French, fix the broken thumbnail in checkout, and write a test for the discount edge case."
Claude reads the whole sentence, finds the independent topics, rewrites each one as a clean task with a proper title, summary and time-bucket estimate.
All tasks land in the backlog, refined and ranked. Select them, hit Start, and Sightdeck AI launches one fresh Claude session per task — they run in parallel.
While you keep dumping new ideas, the Ready-Backlog is already being worked off in parallel. The dashboard streams every status change live across projects — you always see exactly what each agent is doing right now.
Step 2 — watch the run unfold
Once tasks are running, Sightdeck AI shows them as a live lane of connected pill-cards. Serial chains link with arrows, parallel runs stack into their own lanes — colour-coded by status. Glance once and you know exactly where every agent stands.
Step 3 — the moves you'll make every day
When a session stops to ask a question, the task moves into a red-bordered, pulsing column. You see it from across the desk — no Slack ping, no terminal switching.
Multi-select a batch and hit Start. Run them in parallel as one fresh session per task, or chain them serially in a single Claude run — Sightdeck AI picks the right shape based on your selection.
Hit the play button on any task. A fresh Claude CLI session opens in the right directory with the right prompt — no copy-paste, no shell wrangling.
Select a batch of tasks and Sightdeck AI analyses them: which can safely run side by side, which depend on each other and must wait in line. You see the recommendation before you hit Start.
Drop a rough sentence into the quick-add box and press Enter. Sightdeck AI creates the task instantly; Claude rewrites the title, summary and estimate in the background. Long sentences auto-split into independent tasks.
Drag in any number of project XML files. The app watches all of them and gives you a single, sorted task list — grouped by project, status and priority.
Point Sightdeck AI at any URL and every completed task fires a POST with the payload. Slack bot, e-mail relay, deploy trigger, status board — anything that listens.
Top = next to run. Drag tasks up or down to reshape what Claude tackles next. The order drives both single-task launch and bulk start.
Paste an external ticket key or URL. Sightdeck AI fetches the instructions, starts a refinement and works on that ticket for you. Screenshots of tickets work too.
Estimate human effort with a fixed bucket, optionally add Fibonacci story points, and see the real wall-clock time after a task is done. Sanity-check Claude's pace.
Step 4 — 30 seconds to get there
Install the macOS app and open it. Add your project in the sidebar — Sightdeck AI discovers the task file automatically.
Hit Setup → Install in the app. The Sightdeck AI skill lands in ~/.claude/skills/sightdeck-ai. Claude Code picks it up on the next launch — that's the only install you'll ever do.
Open your project in Claude and give it a task. The skill is already active. Status updates, history events and concept docs flow into the dashboard in real time — no terminal switching, no extra setup.
Pick your shape
Park every random idea in the backlog throughout the day. Click Start in the evening and let one Claude session burn through them while you cook dinner.
One XML per client project. Switch between projects in the sidebar — context follows. Spot which clients have agents waiting for your reply.
One pipeline across every agent in flight. See who is waiting on a reply, who is still grinding, and who shipped — without opening N terminals.
Group projects in the sidebar, get a single "needs your feedback" view across the whole portfolio, and never miss an agent that has gone silent.
Good to know
Yes. Sightdeck AI is a layer on top of Anthropic's Claude Code CLI. If you don't have it yet, the app's Setup screen will install it for you in one click.
Locally, in plain XML and Markdown inside your project (.agents/sightdeck/). Sightdeck AI never uploads anything. Commit the files to git if you want them shared; ignore them if you don't.
The desktop app currently ships for macOS (Apple Silicon and Intel). Windows support is in progress. The skill and CLI scripts are pure Python and run anywhere Claude Code does.
Nothing. The XML keeps living in your project. Claude keeps writing to it. Open Sightdeck AI next week and everything is exactly where you left it — only with whatever the agents did in the meantime added on.
Yes. The skill is the contract. Install just the skill into ~/.claude/skills/sightdeck-ai, and Claude will manage tasks in the XML on its own. The dashboard makes it pretty — it isn't required.
Yes — Sightdeck AI fires a webhook POST every time a task moves to done. Point it at a Slack bot, an e-mail relay, a deploy script or anything else with a URL. The payload carries the task id, title, status, run lane and timing — wire it up in minutes, no plugin system needed.
The license is a yearly subscription. One subscription activates the app on up to two machines. All updates released during the active period are included.
Launch incoming
Sightdeck AI launches later this week. The macOS app will be available right here — single binary, no account, no installer drama.
A simple yearly subscription. All updates released while your license is active are included.
Checkout via Paddle. Receipts include VAT for EU customers. We don't store payment details.
Sightdeck AI orchestrates an external AI assistant. AI output is inherently unpredictable and depends heavily on the codebase, the prompt, and the model in use. Every project is different — results will vary, and edits can be wrong, incomplete, or unsafe.
Sightdeck AI is provided as is, without any warranty, express or implied. No guarantees are made about correctness, fitness for a particular purpose, uptime, data integrity, or business outcomes. All use is at your own risk. To the maximum extent permitted by law, no claims, demands, or liabilities — including legal ones — can be derived from anything Sightdeck AI, or the AI it dispatches, does or fails to do. Always review AI-generated changes before committing or shipping them.
Sightdeck AI is an independent product and is not affiliated with, endorsed by, or sponsored by Anthropic or Claude. "Claude" is a trademark of Anthropic, used here only to refer to the third-party AI assistant that Sightdeck AI integrates with.