CREW ONE
Now in private beta

There’s a new shape of company.
It starts with one hire.

Reads your inbox. Sits in your meetings. Watches your data. Closes the loops.

See it run a real week
crew-one.app / chief-of-staff
LIVE
Chief of Staff
Today's briefing
Calendar & travel
Q3 board prep
Hiring · 4 roles
Crews
EAactive
Financethinking
GTMidle
Legalawait

Today's briefing

· Tue 13 May, 08:14
Chief of Staff · Briefing

Three things need you this morning. Finance has a £42k vendor renewal flagged for your approval — Legal already cleared the terms. GTM drafted outreach to the 12 leads from Tuesday's webinar; the framing matches what we used last quarter. And EA is holding the 11:00 with Nadia until you confirm the agenda.

Pull the last three vendor renewals over £30k. Did we negotiate down?
Finance
Thinking
Context
Cited memory
"Renewals over £30k go to Legal first, then Alex for sign-off."
· Set by you, 12 Feb 2026
Today's surface
Inbox3 awaiting
Slack11 mentions
Calendar4 meetings
Work from anywhere
WebWeb
DesktopDesktop
iOSiOS
ChromeChrome
SlackSlack
WhatsAppWhatsApp
Powered by leading AI models
AnthropicAnthropic
GoogleGoogle
OpenAIOpenAI
PerplexityPerplexity
xAIxAI

Crew One routes per task — fast lookups on Haiku, briefings on Opus, search on Perplexity. You see credits, not vendors.

[01] The thesis

The visible parts of a company are getting easy to copy.

Anyone with a credit card can spin up a landing page, an inbox, a brand. What stays hard is the part you can't see — the decisions, the precedents, the muscle memory of how a company actually runs.

WHAT COLLAPSES
Building a product
Reaching customers
Hiring contractors
Producing content
WHAT COMPOUNDS
Knowing what you decided
Knowing why you decided it
Knowing what's worked before
Knowing what to do next

Crew One gives your company a shape that wasn’t possible before AI.

[02] The bottleneck

Every AI conversation starts from zero.

You explain your business to ChatGPT.

Again.

You walk Claude through your customer.

Again.

You re-onboard a new tool with your context.

Again.

That's the bottleneck.

R
Rules
Brittle.

Break the moment context shifts. A new tool, a new vendor, a new exception — and the rule no longer fits. Rebuild from scratch.

D
Decision traces
Compound.

Every decision becomes precedent for the next one. Context drift makes them sharper, not stale. The system gets smarter with every approval.

Example trace

Decision trace

· LAST 24H
14:32
FINANCE· Acme Corp · 20% annual discount approved

Precedent from Q3 2025 renewal; matches Sarah's pricing rule.

APPROVED
[03] Decision traces

Every decision, timestamped and reversible.

Approvals become precedent. Ask 'what did we decide last time?' and get the answer — with rationale, agent, and outcome.

crew-one.app / decision-trace
LIVE

Decision trace

· LAST 24H
08:17
FINANCE· Flagged 4 unused Notion seats for downgrade

Seat-usage export cross-checked against the licence ledger — none active in 90 days. Downgrade queued for approval.

PENDING
08:11
GTM· → Legal · Mercury MSA cross-crew handoff (5-yr survival → 2-yr)

GTM cited Q3 Acme precedent. Same shape — Legal cleared in 13 minutes against the matched template.

APPROVED
08:14
EA· Reschedule Nadia 11:00 → 14:30

14:30 is the only mutual slot before Friday; Nadia's assistant pre-approved this range.

AUTO
08:09
FINANCE· Approve £42k Datadog renewal

Over the £30k threshold. Legal cleared terms 08:02.

PENDING
07:51
GTM· Draft webinar follow-up (12 leads)

Matched Q1 framing per your note 'lead with the customer, not the demo.'

APPROVED
07:33
LEGAL· Counter-sign Mercury mutual NDA

Mercury added a non-standard 5-year survival clause. Pushed back to 2.

APPROVED
Mon
REVOPS· Reject Q2 attribution model rewrite

You have previously rejected first-touch models. Holding to last-touch.

REJECTED
[04] Human-in-the-loop

You stay in control. AI accelerates, doesn't replace.

Approve, edit, or reject any action. Threshold rules learn your preferences — the crew earns autonomy over time.

crew-one.app / approvals
LIVE
Finance · Awaiting you· 2 minOVER THRESHOLD · £30K+

Approve Datadog renewal · £42k / year

Auto-renews Friday 23 May. Legal cleared terms; the 12% YoY increase matches the contract escalator you signed in 2024. Two cheaper alternatives surfaced (Grafana Cloud, Honeycomb) but neither match your APM + RUM coverage.
⌘+↵ to approve
LegalCounter-sign Mercury mutual NDA12 min
GTMSend webinar follow-up to 12 leads47 min
External communications
Outbound email, social posts, customer messages.
Financial transactions
Renewals over threshold, vendor contracts, transfers.
Policy changes
Pricing rules, hiring decisions, brand guidelines.
Data modifications
CRM writes, schema changes, irreversible deletes.
Trust tiers · per connection
Gmail
Drafts sent to you. No outbound send.
● granted
Google Calendar
Drafts events. Confirms after your nod.
● granted
Slack
Posts to you. Drafts to channels.
● granted
Notion
Reads pages. Never writes.
● granted
Stripe
Reads ledger. Never charges.
● granted
HubSpot
Drafts emails. Updates pipeline.
● granted
Linear
Files issues. Triages backlog.
● granted
Quickbooks
Reads books. No edits.
● granted

Per connection, per crew. Tighten the tier on the way in; widen as the crew earns autonomy. Every promotion lands in the trace.

[04.5] Every Surface

Same conversation. Every surface.

Email · Voice · Slack · Web. One thread. One memory. Switch surface mid-sentence — the crew keeps going.

iOS · Voice
LIVE
Chief of Staff
02:14 · in-car · driving
You
What’s the deal with the Mercury MSA?
Email · Gmail
THREAD
Mercury MSA · ready for your eyes
Chief of Staff <crew@yourdomain.com> · 08:11
Hi Alex — GTM passed Mercury to Legal at 07:58. We pushed the 5-year survival clause down to 2 — same shape as the Q3 Acme precedent. Counter-signed copy attached. One question for you below.
Slack · DM
UNREAD
#crew-one3 new
C1
Chief of Staff08:11
Mercury MSA: 5-yr clause pushed to 2. Same precedent as Q3 Acme. Email queued for your approval ↑
AB
You08:12
what’s the deal with the mercury msa
Web · crew-one.app
OPEN
chief-of-staff · mercury msa
⌘K
Picking up where you left off · iOS · 08:12
Chief of Staff · 08:13 →
Full thread here, plus the redline diff vs the original draft. The Acme precedent is cited inline. Approve? ↓
Listening — say a word, type a word, same crew is on it.
01
iOSasked in the car
02
Emailqueued for approval
03
Slackmirrored to your DM
04
Webthread continues here
[05] The crews

Eight teams. One Chief of Staff.

Finance · GTM · Product · Marketing · Legal · RevOps · Ops · EA. Talk to your Chief of Staff. It delegates to the team that owns the work — and remembers what they decide.

Crew Framework ↗
8 crews · 1 memory · live
Mon 20 May · 08:14
ALL ACTIVE
Q3 precedent
Brand voice
Chief of Staff
+ shared memory142 routed
GTMHubSpot
FinanceStripe
LegalDrafts
OpsVendors
RevOpsPipeline
MarketingNotion
ProductLinear
EACalendar
Or compose your own
Investor Relations CrewCustom · built by you
Monthly update, KPI pack, intro pipeline, board-meeting follow-throughs.
BUILT04 Mar
SPECIALISTS4
DECISIONS113

Every team draws from the same memory. Add teams as you grow. Your context follows.

[06] The memory

It remembers everything. So you don’t have to.

Start a conversation on your phone. Continue on your laptop. Reference a decision from six months ago. The crew remembers — across surfaces, across sessions, across teams.

6 months ago
We decided on 20% annual discount precedent.
3 months ago
Sarah led the Series A positioning.
Yesterday
Board prefers the shorter ARR definition.
Pulled precedent from 6 months ago
Chief of Staff · 200ms

The crew pulls precedent like this whenever a decision has one. You see the rationale, the source, and the time it took.

[06.5] The Loop

It learns. Then it learns better.

Tell the crew when something's wrong. It updates its own instructions. When a tool's missing, it builds it. When a run fails, it proposes a fix overnight. The longer you run, the smarter the crew gets — because the loop closes itself.

01Feedback loop

Tell the crew when something's wrong. It updates its own instructions. The class of mistake stops happening.

You teach once. The crew remembers forever.
AB
YouTue 14:23 · #crew-one
you got the discount calculation wrong on Acme — annual prepay nets 12%, not 10. it’s tiered by deal size. update the playbook.
SKILL.md· gtm/discount-policy+4 / −2
Annual prepay: flat 10% across all deals.
+ Annual prepay: tiered.
+ < £50k ARR → 8%
+ £50–250k ARR → 12%
+ > £250k ARR → 15%
Next time · Friday 17:02Mercury quote drafted with tiered prepay (12% — £180k ARR). Correct on first try.
02Tool synthesis

When a tool's missing, it builds it. The first time. So every time after costs nothing.

When a gap shows up, the crew closes it.
22:14
RevOps · agent-007
REASONINGNeed to reconcile Stripe payouts with Bank ledger. Will compose one.No tool exists for this pair.
22:15
Composer
DRAFT · CLIcrew tool new --in stripe.payouts --in bank.ledger --out diff.csvstripe-bank-reconcile --in stripe.payouts --in bank.ledger --out diff.csv
22:16
You
HITL · APPROVEDScope looks right. One-tap approve from the morning queue.
22:17
Tool · stripe-bank-reconcile
FIRST RUNReconciled 1,284 payouts. Found 3 mismatches totalling £412 — flagged to Finance. Tool added to the library.
added
Added to toolkitstripe-bank-reconcile
14 uses since
03Self-healing

When a run fails, it proposes a fix. A second crew reviews. The merge happens overnight. You wake up to a working build.

Failures fix themselves while you sleep.
22:14
Eval · gtm.outreach.v3
FAILUREPersona-match score dropped −0.18 on the founder cohort. Class: prompt drift after model refresh.
22:31
GTM · agent-003
FIX PRRe-anchored persona block to the v2 specimen. +12 / −7 lines. Reasoning + diff attached.
22:39
RevOps · agent-007
PEER REVIEWDiff is scoped, regression suite green. Approved without escalation.
22:48
Crew One
MERGED · NO HITLInside the autonomy envelope for self-healing fixes. Logged to your trace; reversible from the morning queue.
06:12
Eval · gtm.outreach.v3
NEXT-DAY RUNPersona-match score +0.22. Class of failure closed.
morning brief
In your morning brief · 08:00One overnight fix. Logged, reviewed, merged. Score recovered. Nothing for you to do.
[08] How it works

Four stations. Then you’re shipping.

00
Onboard
Day One. Five minutes. First crew live.
Three questions — name, company, role — seed instructions.md (the crew's briefing on you). Connect your inbox + calendar. The EA crew is live by the time you close the wizard.
Day One wizard · 04:12 in
NameAlex Bastien
CompanyCrew One
RoleFounder · CEO
instructions.md seeded · EA crew live
01
Connect
Plug in the surfaces your team already runs on.
Crews come pre-wired for Google Workspace, Slack, Notion, HubSpot, Stripe, Linear, and 40+ more. Most teams finish setup in under 20 minutes.
+ 40 more
02
Delegate
Talk like you'd brief a colleague.
Voice, text, email — Crew One routes to the team that owns the work and asks you only when something earns your attention.
Chief of Staff
Listening
“Prep me for my 2pm A16z.”
03
Compound
Every decision lands in the trace.
Pull precedent by question. Reverse anything. The system gets sharper with every approval.
Decision trace· COMPACT
GTM→ Legal · Mercury MSA cross-crew handoff (5-yr survival → 2-yr)

GTM cited Q3 Acme precedent. Same shape — Legal cleared in 13 minutes against the matched template.

APPROVED
EAReschedule Nadia 11:00 → 14:30

14:30 is the only mutual slot before Friday; Nadia's assistant pre-approved this range.

AUTO
FINANCEApprove £42k Datadog renewal

Over the £30k threshold. Legal cleared terms 08:02.

PENDING
GTMDraft webinar follow-up (12 leads)

Matched Q1 framing per your note 'lead with the customer, not the demo.'

APPROVED
[09] The difference

Most AI remembers context. Crew One remembers decisions.

FEATURE
GENERIC CHAT AI
TASK AUTOMATIONS
CREW ONE
Remembers recent context
Remembers your decisions
Learns from exceptions
Surfaces precedents
Works across 6+ surfaces
Updates its own instructions from your feedback
Builds new tools when it hits a gap
Fixes its own failures overnight
[10] Security

Asks before it acts.

Drafts before sends · Trust tiers per connection · Budgets per run

HITL Gates
Nothing sensitive happens without your approval.
Audit Trail
Every decision logged with rationale.
Data Control
Export or delete your data anytime.
Multi-Auth
Email, Google, GitHub, Passkeys, Phone OTP.
httpOnly Cookies
No tokens in localStorage — immune to XSS.
SOC 2 Type 1Annual independent audit. Report on request.
Pen-testedQuarterly third-party penetration testing.
Audit logEvery action logged with rationale and provenance.

You own your data. Always.

[10.5] The Playbook

Your operating manual. Updated every Monday.

Crew One ingests every decision, meeting, and pattern from the past week and regenerates a living 'How Your Company Works' document. Diff against last week. Searchable. Cite-able. Pump it back in as crew context and the loop tightens.

C1Playbook · v37Week of 18 May
How Crew One works

Your operating manual,
regenerated each Monday.

Table of contents
01How we decide14 sections · 47 precedents
02How we route8 crews · 23 handoff patterns
03How we approve12 thresholds · 6 escalation rules
04How we onboardinstructions.md seed + Day One wizard
05How we deal with exceptions19 named exceptions · 4 open
Generated by Chief of StaffCited 142 decisions · 38 meetings · 12 docs
This week’s diffv37 vs v36 · 5 changes
+0added
~0revised
0obsoleted
+
Added§ 01 How we decide
"Annual prepay discount tiered by deal size: <£50k → 8%, £50–250k → 12%, >£250k → 15%."
From Tue 14:23 · Slack feedback · you
+
Added§ 05 Exceptions
"Mercury-shape contracts: 5-year survival clauses pushed to 2 by default. Cited as Q3 Acme precedent."
From 08:11 · Mercury MSA approval · Legal
+
Added§ 02 How we route
"Cross-crew GTM → Legal handoffs auto-include precedent citations from the past 90 days."
From this week · 4 successful handoffs
~
Revised§ 03 How we approve
"Vendor renewals over £30k → Legal first, then you (was £25k)."
From 12 Feb · your raised threshold
Obsoleted§ 01 How we decide
"First-touch attribution model under review for Q2 rollout."
From Mon · RevOps · rejected by you

The longer you run, the thicker your playbook.
The thicker your playbook, the smarter the crew.

Companies still run on org charts.

Information runs up through middle managers. Decisions run back down. Most of the work is people moving context between apps.

A new shape of company is being built — smaller, faster, recursive. The work happens in loops. Each loop teaches the next. Memory compounds. Headcount stays small because the company doesn’t run on headcount any more.

Crew One is the first hire for that company.

Dogfooding

Crew One runs Crew One. Daily.

The crew on this page runs our own ops — morning briefings, inbox triage, follow-through — with every decision logged. We're user zero, and we hit every rough edge before you do.

Founding 100 · limited to 100 seatsSee founding pricing →
[11] Pricing

Fractional hire value. SaaS pricing.

We're replacing the £5k/month fractional hire, not the ChatGPT subscription.

Starter
£99/ month
For solo founders
1 crew
25 SOPs
Email + web
Community support
Founding 100
£500/ month
Limited — 100 seats
All 8 crews
150+ SOPs
13 domain agents
Shared memory
Slack + WhatsApp
Team
Custom
For growing teams
Everything in Founding 100
Multi-seat
SSO & audit logs
Priority support
Same features. Different usage.
What are credits?
Every plan ships with the same crew, the same memory, the same 150+ SOPs. The difference between tiers is how much you run them — credits cover model usage, voice minutes, and SOP executions. Pay for outcomes, not seats.
Included in all plans
AI Chief of Staff + all 8 specialist teams
All 6 surfaces (Web, Mobile, Slack, WhatsApp, Email, Chrome)
Full persistent memory + decision graphs
All model routing (Haiku, Sonnet, Opus, Perplexity)
Voice input + output
150+ SOPs (standard operating procedures)
Human-in-the-loop approvals
No per-seat pricing — pay for usage, not headcount
[12] FAQ

Questions & answers.

For less than a fractional hire

Stop running your company on tabs.

Start running it on a Chief of Staff.

Or watch a 90-second walkthrough first.