agent session monitoring · for every coding agent your team runs
ZeroShot is agent session monitoring for your team's coding agents. It turns your team's sessions and PR reviews into reusable skills — so agents get it right the first time, any session hands off to any teammate, and skill usage is enforced on every PR.Session monitoring makes teams ship 3.7× faster — on 90% less wasted token spend.
CLI
❯ curl -fsSL https://tryzeroshot.com/install.sh | sh
Works with every coding agent · open-source skills · can run fully locally
zeroshot · daily report
mon 09:00 · posted to #eng
time saved this week9.2 hrs
team efficiency76 / 100 ▲
your rank#2 of 5
vs all teamstop 4%
active sessions
●oauth-rewriteworking
●stripe-webhookneeds approval
●sidebar-redesignwaiting
skills · new / updated3 / 17
repeat work skipped$2,140
money saved this week$9,480
The morning report your team wakes up to — what shipped overnight, what it saved, and which skills to sharpen next.
the receipts · why this exists
"I'm blocked from using agent tools the second I hit my budget. ZeroShot unblocks me — same cap, twice the output. That's the whole pitch."
— Nikhil Shinday, CTO
"One of my engineers spent $40,000 on tokens last month, and I genuinely don't know whether I should stop him."
— Vitaly Gordon, CEO, Faros AI · via TechCrunch
Agents now write a tenth of the world's code, and the spend is getting impossible to justify. The teams that win are the ones getting twice as far on the same budget.
−90%
wasted token spend per task, once your skills kick in
3.7×
faster delivery — agent session monitoring finds what's slowing your team down
2×
output per plan — the same subscription ships twice the work
−67%
PR review time — reviews cite your conventions instead of re-litigating them
└ pilot teams · 90-day window · methodology on request
the racket
Every prompt you type is a first draft.
Every session starts from zero.
Same discoveries. Same dead ends.
Same tokens.
Your coding agents are
ripping you off.
Model providers are not in the business of making your tokens go further. The models get smarter and more expensive every quarter — and your system stays exactly as dumb as the day you installed the CLI. The fix isn't a bigger model. It's a system that learns: your sessions become skills, your skills compound, and the same problem never costs you twice.
It's like the crack-cocaine epidemic — you're kind of beholden to it.
— Chris Reed, Sr. Director of IT Finance, Priceline · via TechCrunch
the proof · same prompt, run twice
Your terminal vs your terminal on ZeroShot.
replayed side by side · watch the meter
no zeroshot · ~/api · tue 9:02am
❯add rate limiting to the /export endpoint
✻I don't see any rate-limiting conventions in this repo. I'll improvise.
⏺Read(src/middleware/…) 14 files
⏺WebSearch("express rate limit best practices")
⏺Write(src/middleware/rateLimit.ts)
✗tests failed · retry 1
✗tests failed · retry 2
⏺Write(src/middleware/rateLimit.ts) "done, I think."
SESSION RECEIPT
tokens74,312
wall clock41 min
retries3
VERDICTRAW-DOGGED
note ─ Sarah shipped the same middleware in March. The agent didn't know. Neither did you.
zeroshot · ~/api · tue 9:02am
❯add rate limiting to the /export endpoint
⏺zeroshot · skill matched: rate-limiting.md
└extracted from Sarah's session · Mar 3 · used 41× since · avg −9.2k tokens
⏺Edit(src/middleware/rateLimit.ts) passes first try
#meanwhile: the other terminal is on retry 2.
❯
SESSION RECEIPT
tokens7,431
wall clock6 min
retries0
VERDICT90% CHEAPER
note ─ Nobody had a meeting about it.
90% cheaper.
Nobody had a meeting about it. Nobody wrote a wiki page that nobody read.
❯ your best engineer's best session, running in everyone's terminal.
[1] · where skills come from
The skills write themselves.
ZeroShot mines every session and PR review for the patterns you keep re-teaching, opens a weekly PR with new skills, ranks them by tokens saved, and retires the ones nobody uses. Every skill is a file — review it, edit it, ship it.
rate-limiting.md
used41×
per run−9.2k tokens
sourcesarah · mar 3
oauth-testing.md
used38×
per run−12k tokens
sourceweekly skill PR
incident-postmortem.md
from6 sessions
per run−5.1k tokens
NEW THIS WEEK
└ start with 13 open-source skills — or steal ours: github.com/buildbetter-app/skills
└ full story → skills
ZeroShot flagged that my Playwright skill was burning about $4k a month. It rebuilt the skill around it — now it's about $200 a month and 10× faster.
— Spencer Shulem, CEO
[2] · session handoff
Hand off the chat.
Not just the code.
Any session resumes on any machine, by any teammate. The agent inherits the reasoning — the decisions, the dead ends, the why — not just the diff. Six months later it still remembers.
sarah's terminal · fri 4:51pm
❯zeroshot handoff --to mike
⏺session packaged · 214 turns · decisions, dead ends & the why included
─── mike's terminal · mon 9:04am ───
❯zeroshot resume sarah/rate-limiting
⏺context loaded in 2s · picking up where sarah left off
✻We chose sliding-window over token-bucket — see incident-42. Continuing.
└ full story → evidence
[3] · observability & enforcement
Observability for you —
and for the whole team.
LOCAL · just you
- ▸Every session you run — tokens, cost, timing, every tool call
- ▸Which of your skills fired, and what each one saved
- ▸Where your own token spend is going
TEAM & ORG
- ▸Every teammate's session — and which skills ran on which PR, not just yours
- ▸What skills are used on what PR, across every repo
- ▸Where tokens are spent across the whole team
- ▸Org-wide — with the guard that blocks a PR when a required skill didn't run
zeroshot report · tue 9:02am
❯what did my agents do last night?
⏺14 sessions · 3 PRs shipped · 2 waiting on approval
└412k tokens (−38% vs last tue) · top skill: rate-limiting.md
└flagged: mike re-derived the auth flow again — skill suggested
─── enforcement ───
❯git push
✗PR #482 blocked — required skill migrations.md didn't run
⏺skill applied · checks pass · PR unblocked
└ full story → observability
[4] · pr reviews
The same review comment,
never typed twice.
ZeroShot drafts your review with /bb-review, citing your team's prior conventions — and every review your seniors write becomes a skill agents apply before the next PR goes up. Teams cut PR review time by 67%.
zeroshot · receipt
PR #412 · api/rate-limit-export
session14m 22s
tokens48,210
run cost$2.10
skills applied
✓/payments-reviewsaved $6.40
✓/migration-guardsaved $4.10
✓/bb-reviewsaved $7.90
✗/export-perfnot run · $3.20 missed
reviewed14 files · 320 lines
flagged2 issues
total saved$18.40
One receipt per PR — what it cost, which skills fired, and what each one saved. The skill that didn't run is the one your team fixes next.
pr review · #3308
❯draft my review with /bb-review
⏺posted to PR #3308 · 3 suggestions, all citing your team's prior conventions
└conversation-table convention · set by maya in PR #408
─── six weeks later ───
⏺zeroshot · nikhil's retry comment on PR #408 → retry-policy.md
✻applied before the PR went up. nobody had to type it again.
└ full story → skills
The only way I let a non-engineer ship to production is if they're using ZeroShot. It works that well.
— Adam Stanford, CPO
❯ ok. give it to me.⏺ two ways in:
