KnowSizePriceLand the work.
sleds is the layer between the people who ask for work and the agents and engineers who do it. Every ask is sized before anyone touches it, put on the sled with what it is worth, dispatched to the seat that fits, and receipted when it lands — written back to Linear, Jira and GitHub in their own words.
| # | ask | size | band | conf | worth | who | filed by |
|---|---|---|---|---|---|---|---|
| ENG-482 | SSO redirect loop on the Acme tenant | M | $4–11 · 20–55 min | 81% | high · Acme renewal | Jake | Samantha |
| ENG-479 | Export CSV respects the saved filter | S | $1–3 · 6–14 min | 90% | medium | overnight | Jake |
| ENG-471 | Billing page: proration shown before confirm | L | $14–38 · 1–3 h | 64% | high · churn | Dana | John |
| ENG-468 | Rate-limit headers on the public API | M | $5–12 · 25–60 min | 77% | — ask the filer | unassigned | Samantha |
| ENG-466 | Empty state for a new workspace | S | $1–2 · 5–10 min | 93% | low | overnight | Samantha |
| ENG-461 | Webhook retries with backoff | M | $6.10 · 38 min actual | — | medium | Jake | John |
size M $4–11 20–55 min 81% basis 112 rows · auth · this repo fits Fable (band) · Sonnet (+40% time, ⅕ cost) · Haiku does not fit worth high · Acme renewal (stated by Samantha) cap $20 · nothing bills past it suggest now · Jake · 62% of the weekly bar · resets Thu 9:00 written to ENG-482 · estimate field · comment
Wall: usage limit reached. Session snapshotted.
Provider: Weekly limit reached. Resets Thursday 9:00 am.
Remaining work: $4.10–9.30 · 20–55 min · 82%
1 Continue now platform key · allowance left $38.00
2 Hold until reset free · resumes 9:00 am EDT · saves ≈ $6.70 vs continuing now
3 Run overnight batch tier · ½ price
4 Switch model sonnet 5 · platform key · $0.90–2.10 · same time · rest of session
5 Hand off to Dana (idle 42m) · Slack + /sled pickup
6 Defer to reset runs itself 9:00 am EDT · same model · full price · $4.10–9.30
Reply /sled 1, /sled 2, /sled 3, /sled 4, /sled 5 or /sled 6. wall jd7k2m9qa41p
> /sled 1
Continuing on the platform key. Worth high · policy allows · $4.10–9.30 against $38.00. Logged.SLEDS-Source: ENG-482 · brief · linear Worth: high · Acme renewal (as she said) SLEDS-Estimate: $4–11 · 20–55 min · 81% · cap $20 SLEDS-Seat: overnight · batch · fable-5.1 SLEDS-Actual: 31 min · $5.20 SLEDS-Attest: ed25519 · 2026-09-14T06:41Z · verified → ENG-482 moved to Done · receipt sent to Samantha · the next size is a little better
Work arrives faster than anyone has ever asked for it. Almost none of it is sized.
Everyone in the company has Claude or ChatGPT now and is making things all day: specs, prototypes, a customer's ask in a markdown file. It lands on engineering unsized, unvalued and unpriced. The agents make the doing cheap. Nothing between the ask and the merge knows what any of it costs or is worth — so nothing can decide what runs now, what runs overnight at half price, what goes to a teammate, and what waits.
One afternoon on the sled.
Samantha asks from her Claude.
No ticket, no login. She says "send this to engineering" with the spec she just wrote. Her Claude, running our instructions, asks the two things the spec did not answer — what does done look like, and what is the smallest version worth doing — and plays the brief back. She says yes.
Sized from the team's own record.
A letter for the glance, a band behind it: $4–11, 20–55 minutes, 81% confidence, from 112 landed rows in this repo's auth area. Which model fits comes from first-try pass and rework rates, not a price list. Written to ENG-482 in Linear's own estimate field.
size M $4–11 20–55 min 81% basis 112 rows · auth · this repo fits Fable (band) · Sonnet (+40% time, ⅕ cost) worth high · Acme renewal (stated by Samantha) cap $20 · nothing bills past it written to ENG-482 · estimate field · comment
On the sled, with worth beside it.
The row sits in the priced, staged queue with everything else the company asked for. John, the lead, sees worth and size side by side, adjusts worth on two rows, and never sees a per-person number. Policy dispatches: Jake owns SSO and has headroom this week.
| ask | size | worth | suggest |
|---|---|---|---|
| SSO redirect loop · Acme | M | high | now · Jake · headroom |
| Rate-limit headers · API | M | — | hold · ask the filer |
| Export CSV filter | S | medium | overnight · ½ price |
| Empty state · new workspace | S | low | overnight · ½ price |
Jake hits the wall. The exits were priced already.
At 2pm his five-hour bar is at 94%. Instead of a dead terminal he gets six exits, each with a cost: continue on the platform key within his allowance, hold until the reset, run it overnight at half price, switch model, hand off to Dana, or defer so it runs itself at the reset. He picks three and keeps building.
Wall: usage limit reached. Session snapshotted.
Provider: Weekly limit reached. Resets Thursday 9:00 am.
Remaining work: $4.10–9.30 · 20–55 min · 82%
1 Continue now platform key · allowance left $38.00
2 Hold until reset free · resumes 9:00 am EDT · saves ≈ $6.70 vs continuing now
3 Run overnight batch tier · ½ price
4 Switch model sonnet 5 · platform key · $0.90–2.10 · same time · rest of session
5 Hand off to Dana (idle 42m) · Slack + /sled pickup
6 Defer to reset runs itself 9:00 am EDT · same model · full price · $4.10–9.30
Reply /sled 1, /sled 2, /sled 3, /sled 4, /sled 5 or /sled 6. wall jd7k2m9qa41p
> /sled 3
Running overnight. Capped $20 · batch tier · ½ price. Delivered by 6:20 am; /sled jobs to watch.It lands with a receipt.
The overnight run merges at 06:41. The PR body carries the record: size against actual, who built it, on what, for how much, signed with the workspace key. Linear moves ENG-482 to Done with the numbers on it. Samantha's Claude gets the receipt in her language.
SLEDS-Source: ENG-482 · brief · linear Worth: high · Acme renewal SLEDS-Estimate: $4–11 · 20–55 min · 81% · cap $20 SLEDS-Seat: overnight · batch · fable-5.1 SLEDS-Actual: 31 min · $5.20 SLEDS-Attest: ed25519 · 2026-09-14T06:41Z · verified → ENG-482 moved to Done · receipt sent to Samantha
Friday, it rolls up. Nobody is counted.
What landed, what it was worth, what it cost, where the pile-up is. By area, never by person. The books are a byproduct of engineers routing their own work.
This week on the sled illustrative landed 47 rows · $212k worth · $1,140 cost in band 78% of sizes overnight 19 rows at half price · $71 saved walls 12 hit · 12 priced · 0 waited review billing is the queue · 2 days median people not counted
Nobody opens sleds. It shows up where they already are.
ENG-482 · M · $4–11 · cap $20
resets thu 9:00 · $38 allowance
cmux · danaEveryone who touches the work. Priced for none of them to be graded.
Intake
File from the AI you already use. Five questions at most, then a brief, then a row. No seat, no login. Free for everyone, forever.
Intake →The sled
Every ask, sized and valued, staged to go. A letter, a band, a worth, a seat — and a receipt when it lands.
The sled →Engineers
Your laptop, your install, your numbers. The best way through your list; a wall that is an exit. Nothing here grades you.
Engineers →Leads
Not the spread of spend — the flow of work. Value in flight, landed, waiting. Cost fourth. By area, never by person.
Leads →The stack you already run.
sleds sits between the tracker and the harnesses. It reads the seats, reads the tracker, listens to the harness, and writes three things: the size and actual to the tracker, the dispatch decision to the seat, and the event to wherever the person is.
Status line, prompt hook, /sled, the exits before the next prompt.
Through the proxy on the workspace's own OpenAI key.
Background agents as an executor; the seat's bars as a rail.
The coding agent as an executor; credits as a rail.
Sized on open; estimate, worth and actual written to the issue.
The enterprise door. Its "assign to an agent" becomes an input to dispatch.
Every PR sized on open, receipted on merge from the SLEDS-* lines.
Hand-offs, walls and digests. Aggregates; never a per-person list.
mcp.sleds.ai/mcp. intake, estimate, dispatch, status, hand off.
The same MCP server through its connector slot.
The seat line per pane, a ring at the wall, the sidebar block.
The long tail and the enterprise cloud, as rails.
One connector. Every verb where the ask was written.
Use the AI you already have. sleds is the record underneath.
Add the connector once. mcp.sleds.ai/mcp
The same server through its connector slot.
curl -fsSL sleds.ai/install | sh
Background agents as a seat. Next.
Day one, the tracker gets better tickets. Day thirty, the sled is the record.
sleds writes the estimate and the actual to Linear, Jira and GitHub in the fields they already have, and replaces nothing. By day thirty it has seen every piece of work start, priced it, and signed what came back — more than the tracker ever saw. The tracker becomes a view a growing share of the team stops opening. We keep it in sync as long as anyone wants it.
Estimate and worth on open, actual and receipt on merge, in the fields the tracker already shows.
Every PR sized on open and recorded on merge. The parser reads every shape the line has ever taken.
The thread hears what landed, what is waiting, what hit a wall. Aggregates, never a list of people.
Plane, YouTrack, GitLab as mirrors too. Or the whole record in your cloud, on the Enterprise plan.