what’s in it
You don’t have to watch it work.
It runs the agents, keeps track of them, and tells you when one of them needs you.
summaries
instead of: reading the whole log
Agents provide simple summaries.
voice
instead of: picking a model and opening a terminal
Just say what you want done.
quotas
Claude Code
at the limit
Gemini
room to spare
moved over · kept its place
instead of: waiting for the window to reset
Hitting a limit doesn’t stop the work.
how a job moves
plans it
Turns your sentence into steps
researches it
Reads the codebase before anyone writes a line
designs it · claude design
Publishes a real page you can look at
hands it off
One doc, written for the implementer
implements it
Builds from the handoff
validates it
Drives the app and checks the screen
instead of: running each agent yourself
One request goes through the whole crew.
instead of: being the test step
It checks its own work, so you don’t have to.
fewer round trips · never a cached frame
every agent has eyes
rae
reads back her own page
Filter works. The timeline overflows at 1280 wide. Fixing it, then I’ll look again.
screenshot · shot-14.png
console · 0 errors
what comes back
A response you can understand.
The same report, drawn. Read in a glance, not a scroll.
what every other agent hands you
## Build time by stage Measured across three runs each. | stage | debug | release | |--------|-------|---------| | cold | 88 ms | 188 ms | | warm | 41 ms | 91 ms | | cached | 12 ms | 32 ms | - Cold dominates in both configurations. - Release is roughly 2.1x debug throughout. - Warm is about half of cold; cached is negligible. Summary: the cold path is where the time goes. Release adds a constant factor on top rather than changing the shape.
what Wispfield hands you
important
Cold builds are where the time goes. Release adds a constant factor on top, it does not change the shape.
188 ms
cold, release
2.1×
release vs debug
41 ms
warm, debug
12 ms
cached, debug
Build time by stage
debug
release
200
150
100
50
0
cold
warm
cached
milliseconds
No setup, no prompting tricks. It is the markdown agents already write.
instead of: staring at the screen to see what finished
She tells you what matters, and only that.
finished, stuck, or needs you · from another room
she talks
Rae finished. The session store rewrite passes the field suite.
14:02
Cole needs you: keep the legacy adapter, or drop it?
14:09
spoken aloud · one line each
instead of: reading through pages of results in the morning
Ask her where you left off.
what needs you comes first
coming back to it
you
“Where did we leave off?”
Six looms on Wispfield.
Cole, waiting on your answer since 9:12
Nia, quiet for 40 minutes on the migration audit
Rae, finished the session store rewrite
Zoe, Finn and Mira are on the field too. Ask me about any of them.
the rest of the day
Smaller things that add up.
After a week you can’t live without them.
Command palette
Ctrl+K
Halt the field
Halt this loom
Everything the app does, behind Ctrl+K.
Asks, then waits
Build it now, or save it for later?
Build
Shelf
It asks on the card and holds its place until you answer.
Gremlin mode
on · free rein · as many helpers as it takes
Usage resets soon? Flip it on and let them burn down the list.
Halt and resume
back
One key stops every agent. Resume picks up the same task.
Color by work
Coding, researching, writing. Each its own tint.
Live preview
A browser on the field shows the app changing as they work.
keys stay on your machine · windows 10 & 11
Get it →instead of: manually triggering agents
Put it on a schedule.
runs with wispfield closed
what runs itself
weekday mornings · 07:00
Read last night’s failing tests, investigate, and fix the issue
Thu 07:00
finished
4m 12s
Wed 07:00
finished
3m 48s
Tue 07:00
skipped
no failing tests last night
bring the plans you already pay for
coming soon
You pay us nothing. It runs on the subscriptions and keys you already have.
Meet Loomi.
Download for Windows
windows 10 & 117-day free trialuse the model plans you already pay for