Anyone with an idea and a Claude tab can put work on the sled.
Most people in a company never had a Jira seat and never will. They have Claude or ChatGPT and they are making things all day: a spec, a prototype, a customer's ask, a markdown file. Intake is the front door for them. From the conversation they already have open, they say what they want and what it is worth, and it lands on the sled sized, in the tracker's own words, with a receipt when it comes back.
Your AI asks the questions a good engineer would ask. Then it files.
Say it, attach it, link it.
A paragraph, a spec, a Loom, a Figma frame, a screenshot of a competitor. Your AI reads it first and only asks what it cannot answer from what you sent.
Done, and the smallest version.
The two questions that move a size the most. Acceptance without calling it that; the wedge that usually halves the estimate.
The brief, played back. You say yes.
Until you do, nothing is filed anywhere. The draft costs nothing and expires in a week.
On the sled, in the tracker, sized.
One paragraph back: the ticket, the size, the worth you stated, and a suggestion. Your name is the reporter. Nothing is filed as sleds.
The receipt, in your language.
What shipped, size against actual, worth as you stated it. Not a diff. A week later you can ask "what happened with the Acme thing" and get the same answer.
Who builds is changing. We serve the seam.
Business users are moving toward engineering: writing the spec, building the prototype in Make or v0 or Bolt, filing the ask with the repro. Engineers are not moving the other way. The builder tools make the business user productive on their own; the coding agents make engineers productive on theirs. The seam between them is where the cost and the hand-off are, and it is where nothing sits.
Intake is what sits there. Someone asks the questions a good engineer would ask; someone turns the answers into a shape an engineer can size without losing the why; and the engineer who takes it sees that why first, because understanding the work is the part of the job that grows as the typing part shrinks.
The why travels with the row.
Stored on the row, rendered into the tracker's issue in the tracker's own words, shown to the engineer at pickup, and returned to you with the receipt. Every field says where it came from — your words, or the artifact. An engineer never picks up a row without it; a brief without a why is a draft.
Two verbs close the loop without either person leaving their surface. Ask the filer: one question from the engineer's terminal goes into your thread; your AI answers from context if you already said it, and asks you only when it cannot. The receipt: back to your thread when it lands.
- ask
- Fix the SSO redirect loop for Acme before their renewal call.
- why · from the spec
- Acme's admins can't sign in from the new tenant. Renewal is the 24th.
- done · from Samantha
- An Acme admin signs in once and lands on the dashboard.
- smallest · from Samantha
- Just the redirect. The audit-log entry can wait.
- touches · from the spec
- tenant config · the auth redirect · Acme's setup
- sources
- spec.md · Make prototype · Loom (2:10)
Wherever it starts, it lives in one place.
Claude or ChatGPT with the connector added once. "Send this to engineering" with a file, a link or a paragraph. The primary door.
/sled intake — an engineer filing for themselves or a teammate. Same tool, same questions.
An issue labeled intake in Linear or Jira becomes a thread; the brief lands back as the body.
Message the sleds app or /sled intake …. Same thread, in a DM.
sleds.ai/i/… for the person with a Figma link and a browser. Opens the thread by email.
Add one connector. Every verb, in the conversation.
Add mcp.sleds.ai/mcp as a connector in Claude, Cowork or Claude Design and sign in with your workspace account. From then on the same verbs the engineer has in the terminal answer in your conversation. A Cowork plugin bundles the connector, the "send to engineering" skill and the spec template, so the business side installs in one line too.
Your AI runs the conversation under instructions we serve it. sleds enforces the parts that matter at the boundary — the brief is a schema, worth is marked by who said it, the filer is the signed-in identity, nothing is written before "yes" — so it works the same whichever AI is on the other end.
| say | get back |
|---|---|
| intake {text, links, file} | The questions, then the brief for confirmation. On "yes", a row. |
| estimate {spec} | A band, a fit, a confidence, a cap. Recorded. |
| dispatch {spec, worth, links} | A ticket in the tracker, priced, with a suggestion. One paragraph. |
| status {ticket} | Who has it, what it has cost against the band, what is waiting on you. |
| handoff {task, to} | Parked for a named teammate or buddy, with the brief. |
| pickup {id} | Work someone sent you, back verbatim, running on your seat. |
| usage | Your meter: bars, credits, the reset, what is queued. Yours in full. |
| help {word} | The whole table, filtered. Chat, terminal and app never disagree. |
| rule | what it means |
|---|---|
| Asks; never decides | You confirm the brief; the lead adjusts worth on the sled; the policy dispatches. Nothing is written to the tracker before "yes". |
| Worth is asked, never guessed | Blank stays blank. A missing worth is never a low. |
| You stay the reporter | Your name on the ticket, your thread for questions, your receipt at the end. Nothing is filed as sleds, nothing as the model. |
| The why travels with the row | An engineer never picks up a row without the brief. |
| Free for everyone | No seat, no login to file. A seat is needed to take work, not to ask for it. |
| Nobody gets graded | No counts of who files, who answers, or how long they take — to anyone but themselves. |