Everything sleds does, in the words you'll use.
sleds answers in three places: the conversation (Claude, ChatGPT, Cowork), the terminal (Claude Code, Codex, Cursor), and the board. The verbs are the same everywhere, generated from one file, so chat, terminal and app never disagree.
Three ways in.
1. In Claude or ChatGPT, add a connector with https://mcp.sleds.ai/mcp.
2. Sign in with your work email; a link arrives in your inbox.
3. Say "send this to engineering" with a file, a link or a paragraph.
1. curl -fsSL sleds.ai/install | sh
2. sleds init (the full path the installer prints, until a new shell) signs you in, lists what it will change and asks Go ahead? [y/N]. --project: this repo only. --undo: take it out.
3. /sled estimate before you build; the actual closes it on merge.
A token is for a headless machine (CI, a box): sleds init --yes --token …. Never paste one into a chat.
1. Sign in at app.sleds.ai with the 6-digit code.
2. Install the GitHub App and Slack; add people as members or seats.
3. Set the four policy numbers: ceiling, allowance, threshold, default at the wall.
Say it. Get back a paragraph.
| say | get back |
|---|---|
| intake {text, links, file} | The questions, then the brief for confirmation. On "yes", a row. |
| touches {question} | What it touches in the product map, and what on the board collides. |
| attach {row, file} | A spec, prototype or screenshot on the row. |
| estimate {spec} | A band, a fit, a confidence, a cap. Recorded. |
| dispatch {spec, worth, links} | A row on the board, priced, with a suggestion. |
| status {row} | Who has it, what it has cost against the band, what is waiting on you. |
| answer {question} | Your reply to engineering, on the record. |
| 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. |
| help {word} | The whole table, filtered. |
Same verbs, three surfaces.
| verb | terminal | claude code | what it does |
|---|---|---|---|
| intake | — | /sled intake | The questions, the brief, the row. |
| estimate | sleds spec-event on save | /sled estimate [file] | A band, a fit, a confidence, a cap. Recorded; the actual closes it. |
| status | sleds status | the status line | Bars, credits, the binding reset, what is queued. |
| jobs · handoffs | sleds handoffs | /sled jobs | Overnight jobs and the team's open hand-offs, yours first. |
| handoff | sleds handoff | /sled handoff · the wall | Park a brief for a teammate or buddy, with the snapshot. |
| pickup | sleds pickup <id> | /sled pickup <id> | Continue a parked task or walled session on your seat. |
| withdraw | sleds withdraw <id> | — | Take back a hand-off nobody picked up. |
| defer | sleds defer [+2h | iso] | the wall · hold until reset | The musher resumes the session at the reset. |
| overnight · resume | — | /sled overnight · /sled resume <id> | Sized, capped, batch tier; the result pulled back in. |
| verify | sleds verify <jobId | attest line> | — | Ed25519 against the workspace's public key, locally. |
| help | sleds help [word] | — | The whole table, filtered. |
| off · undo | sleds off · sleds init --undo [--project] | — | Back to your seat directly; or take out what init wrote, asked first. The token stays. |
Common questions.
Do business users need a seat?
No. A member of the workspace can file, answer questions and check status from Claude or ChatGPT. Seats are for the people who take work.
Do we still need a tracker?
You can keep yours, or not. The board holds every request with its brief, size, worth, seat and receipt, and stays in step with Jira or Linear both ways for as long as you run them. If you later want to run on sleds alone, they become read-only history.
How is a size worked out?
From your own landed rows of the same task type, repo and model. A letter for the glance, a cost band, a time band, which models fit, a confidence and a cap. Never a point figure.
What happens at the wall?
At 90% the seat tells you. At 100% the session is snapshotted and you get six priced exits; if nobody answers, the workspace policy chooses and the audit log says so.
What do you keep?
The numbers, never the prompts or the code. Three features carry content and each is yours to turn off. See Trust.
Can't find it?
Ask your AI "sled: help", or request an invite and we'll walk your team through it.