CloudSwarm · launch an agent. automate your business. · sovereign Canadian mode available
Launch an agent. Automate your business.
Point-and-click agent builder. Pre-wired skills for the tools you already use. Approvals and audit built in, so you actually sleep at night.
Tell CloudSwarm what you want done in plain English. It picks the right skills, asks for the creds it needs, builds and previews the work, then waits for your green light before it acts. Runs on the R1 runtime. Sovereign Canadian mode when jurisdiction matters.
→ research 50 dentists and send booking offers
→ read my inboxes and track invoices daily
→ ship a React Native app end to end
→ reconcile Stripe to QuickBooks each Friday
agents launched:4,812
skills in marketplace:218
integrations:94
hours saved / mo:{{TBD}}
regions:us · eu · asia · bc
brand:CloudSwarm
Live runtime
region · us-east-2 · bc-van-1
agents running
247
steps / min
1,284
outcomes · last 24hrolling
what it does · 01
Give it a job. It shows up every day and does the job.
Agents here aren't chat windows. They're small, scheduled workers with skills, budgets, and a paper trail — doing the repetitive work of your week so you can do the rest.
📬
Inbox, already sorted
Reads new email every 15 minutes, drafts replies, files receipts, flags anything that actually needs you.
📊
Monday metrics, auto
Pulls the week's numbers from Stripe, Posthog and your sheet; writes the Slack standup your team actually reads.
🔎
Competitor watch
Scans 40 sites + RSS + product changelogs, summarises what moved this week, emails a digest with links.
🧾
Expense janitor
Watches a Gmail label, extracts amounts, categorises, posts to Xero. No more end-of-month archaeology.
📝
Meeting follow-ups
Reads the transcript, drafts the recap, pulls out action items, creates the Linear tickets with owners.
🛎️
Support triage
First-pass on new Intercom tickets: tags, priority, a suggested reply. Your team reviews, hits send.
how it works · 02
Three steps. No prompt engineering degree required.
You describe the outcome. We wire up the skills, the schedule, and the guardrails. You watch the first run, nudge it, and hand it the keys.
Describe the outcome
"Every Monday at 8am, summarise last week's revenue and post it to #growth." Plain language is fine.
plain englishor pick a template
Plug in skills
Click on the connectors it needs — Stripe, Gmail, Slack, your sheet. Set a monthly budget. Approve the permissions.
180+ skillsOAuthbudget cap
Watch it run
Every run shows up as a log entry with inputs, outputs, cost, and the exact tool calls. Pause, tweak, or fire it again.
full audit trailrollback
the dashboard · 03
A home for every agent you launch.
One place to see what's running, what it cost, and what it did. Not a wall of logs — a calm list of workers showing up for you.
📬
Inbox triage
runs every 15 min · 4 skills · $0.12 avg
● running
emails handled · 7d
1,284
↑ 12% vs last week
drafts accepted
92%
steady
spend · 7d
$8.40
under budget
recent runs
2m ago
Replied to 4 emails · filed 2 receipts · flagged 1 for review
ok · $0.14
17m ago
Replied to 2 emails · unsubscribed from 3 lists
ok · $0.09
32m ago
Paused for your review — "Contract from Apex Legal"
waiting
build · 01 / 07
Point-and-click agent builder. Three minutes to launch.
Pick a template or start blank. Name it. Give it a goal. Turn on the skills it needs. Set a budget. Pick a schedule. Hit launch. Everything you toggle here is a real capability under the hood.
templates6 shown · see all 40+
configureunsaved · configuring
launch · 02 / 07
Tell it what you want. It does the rest. You approve each step.
Three real jobs from real customers, walked end to end. The agent figures out what it needs, asks for creds when it doesn't have them, drafts the work, shows you a preview, waits for approval, and only then acts. Every step is auditable.
agent · outreach-01
step 1 of 8
reply or tweak…
⏎ approve · ⇧⏎ tweak
what you see · 03 / 07
One pane. Every agent. Every workspace. Every policy.
This is your CloudSwarm control plane once a few agents are running. Four panels, live. Click into any row to inspect. The whole operation is legible from one view.
workspaces · 7
fleet · 6 running
policy stream
approvals · 4 pending
how trust expands · 04 / 07
Five trust levels. Agents earn their way up.
Every new agent starts at L0: read-only, approval on every action. Trust expands as evidence accumulates. You decide when to raise the cap. An agent never promotes itself.
five levels · click any card to inspect
a typical tuesday · 04
Here's what a day looks like once they're running.
Not a flashy demo — just the calm, small stuff that adds up. You get coffee; they've already handled the first three things on your list.
06:42
🌅
Monday metrics · scheduled run
Pulled revenue from Stripe, sessions from Posthog, signups from the sheet. Drafted the #growth post.
08:03
📊
Monday metrics · posted to Slack
Revenue up 14% WoW, signup conversion steady, one anomaly flagged: Brazil traffic doubled overnight — "worth a look."
08:11
☕
You · reading
Thumbs up on the post. Reply: "nice catch on Brazil — dig into it?"
08:14
🔎
Monday metrics · follow-up
Ran a referrer breakdown, found a TikTok creator who posted about you. Added to the weekly digest.
10:30
📬
Inbox triage · ongoing
Handled 14 emails this morning: 9 replied, 3 filed, 2 unsubscribed. Paused on a contract from Apex Legal — waiting for you.
11:45
✅
You · reviewed and approved
Skimmed the Apex contract in 90 seconds. Approved the suggested reply. Unpaused.
13:02
🛎️
Support triage · first-pass
Tagged 8 new tickets. Drafted replies for 5. Escalated 1 ("refund + compliance") to your queue with a summary.
15:30
📝
Meeting recaps · Q3 planning call
Read the transcript. Drafted a 6-bullet recap. Created 4 Linear tickets with the owners you mentioned.
17:48
🧾
Expense janitor · end of day
12 receipts filed. Categorised, amounts verified. Posted to Xero. Monthly spend: on track.
19:11
🌙
Competitor watch · evening scan
Scanned 42 sources. Tomorrow's digest is drafted and queued for 07:00.
governance, live · 05 / 07
Cedar policies. Evaluated inline. Explainable.
Toggle policies on and off. Run a sample agent action against them. See which matched, which denied, which advised. Real governance, not prompt gatekeeping.
policy matrix · toggle to enable
Policies are Cedar. Real engine. Every decision auditable.
sample actions · click run to evaluate
multi-day work · 06 / 07
Three days. Four approvals. One provider outage survived.
Scrub a real three-day outreach workflow. Four approval gates, a provider failover mid-day-two, resume from WAL after a restart on day three. This is what durable means.
t = 00:00 / 72:00
0h12h24h36h48h60h72h
running
approval wait
failover
recovery
complete
current state
running
summary
Three days. Four approvals. One provider outage. One system restart. One workflow.
Illustrative workflow. Real jobs vary.
what it costs · 05
Your monthly bill looks like a phone bill. Not a cloud invoice.
One line per agent. What it did, what it cost. No surprise egress charges, no "committed use discounts" math, no 47-page PDF from finance.
October · workspace "acme-hq"
Your agents, this month
$142.08
of $250 budget
📬
Inbox triage2,148 emails handled · 92% drafts accepted
2,148 × $0.014
$30.07
📊
Monday metrics4 reports · 12 follow-ups
16 runs × $0.38
$6.08
🔎
Competitor watch31 digests · 1,240 sources scanned
31 runs × $1.22
$37.82
🛎️
Support triage612 tickets tagged · 340 drafts
612 × $0.06
$36.72
📝
Meeting recaps28 calls · 104 tickets created
28 × $0.71
$19.88
🧾
Expense janitor186 receipts filed to Xero
186 × $0.064
$11.91
Total · October
$142.08
✓
Every line is clickable. Drill in to see every run, the exact prompt, the tool calls, and the output. Nothing is a black box — including the bill.
choose your boundary · 07 / 07
Same product. Two boundaries. One for your regulated work.
CloudSwarm is multi-tenant, US/EU/Asia, standard enterprise posture. CanadianClaw is single-tenant BC Canada, enforceable in BC courts under a Canadian-entity MSA. Pick the boundary that fits the contract you're signing.
Affects hero brand and Demo 1 workspace filter.
dimension
CloudSwarm · multi-tenant
CanadianClaw · sovereign
Most workloads belong here.
Regulated workloads belong here.
what's rare here
Five things that don't exist in the rest of the market.
Not features. Architectural bets that only pay off because CloudSwarm is a GUI-first control plane built on a serious runtime (R1) with real governance underneath.
01
The builder does the hard parts for you.
Most agent platforms show you a blank prompt box and wish you luck. CloudSwarm asks what you want to do, picks the skills, requests the creds it needs, drafts the work, and stops at every risky step to check with you. You are the editor, not the operator.
02
Durability is structural, not a retry loop.
A WAL-backed event bus reconstructs an agent's entire state from the log. A provider outage mid-day-two does not reset day one. The agent picks up where it left off with full context.
03
Progressive trust is a product mechanic, not a config flag.
Agents literally cannot act at L3 if they haven't accumulated L2 evidence. Autonomy is a state the system arrives at because the evidence says it should, not a switch you flip.
04
Built-in skills for the tools you actually use.
RelayGate, Actium, Actium Studio, TrueCom, DeepTap, Veritize ship as first-party skills. Gmail, Stripe, QuickBooks, Calendly, Cloudflare, GoDaddy, Namecheap, Fly.io, Vercel, Expo, Google Sheets come out of the box. The agent uses them the same way you would: by logging in once and following the UI.
05
Two brands, one substrate, real sovereignty.
CanadianClaw is the same product under a Canadian-entity MSA enforceable in BC courts, running on the Good Ventures Lab BC-Canadian inference substrate. The no-logging posture is contractual, not just architectural.
Each claim corresponds to a live product mechanic. Security posture details on /security.
in the portfolio
CloudSwarm is the control plane. These are the pieces it conducts.
Every sibling ships as a first-party skill you can turn on from the builder. Use CloudSwarm alone or with any subset.
CloudSwarm+R1
R1 is the runtime under every CloudSwarm agent: harness-driven, STOKE-audited, cross-family reviewed.
r1.run ↗CloudSwarm+Actium
Agency workflows ship as installable skills. Let Actium handle outreach, content, and campaign ops inside your agent.
actium.app ↗CloudSwarm+DeepTap
Grounded answers instead of plausible ones. DeepTap feeds every agent session fresh facts from your corpus and the web.
deeptap.ai ↗CloudSwarm+TrueCom
Commerce settled per action. TrueCom signs every outbound payment across ten rails.
truecom.ai ↗CloudSwarm+RelayGate
Traffic governed at the edge. RelayGate enforces flow policy on every inbound and outbound request.
relay.one ↗CloudSwarm+Veritize
Every agent output verified and drift-tracked. Artifact-level checks as a gate condition in your workflow.
veritize.app ↗CloudSwarm is the conductor. Each sibling runs standalone. Compose on your timeline.
what people say · 06
What changes after a month.
Not benchmarks. Just the mundane, human stuff people notice once a few agents have been running in the background for a while.
"I stopped opening Stripe on Mondays. The agent posts the numbers before I've made coffee. I didn't realise how much of my headspace that weekly ritual was taking."
MR
~4 hrs/week back
"It's the first 'AI tool' I've used where the receipts are legible. Every run shows cost, inputs, outputs. I can show my CFO exactly what we spent $180 on last month."
DT
audit-friendly
"We gave the support triage agent one week on 'draft only, don't send.' It was right ~90% of the time. Now it sends. My team reviews what it escalates."
PK
92% first-pass accuracy
"I expected to write a bunch of prompts. I wrote a sentence, toggled four skills, hit launch. It's the first no-code thing that didn't feel like a toy in three years."
LW
3 agents in day one
"The 'pause for approval' feature sold it for me. It does the 95% of work that's safe, then stops on the 5% that needs a human. That's exactly how I'd delegate to a person."
JK
0 "oops" moments
"Compliance signed off in a week because the audit log is just… there. Every tool call, every input, every output, every cost. No other vendor gave us that."
SN
SOC-2 ready
pricing
Three tiers. Plus a sovereign option when jurisdiction matters.
Start free. Upgrade when you ship real work. Full comparison on /pricing.
Starter
For solo operators and small teams launching their first agents.
free · forever
- 1 workspace, 3 running agents
- Core skill catalog, builder GUI
- Community support
Most popular
Team
For teams running agents in production every day.
{{TBD-team-price}}/mo per workspace
- Unlimited workspaces, up to 200 running agents
- Full skill marketplace, Cedar policies, vault, approvals
- Priority support, quarterly review
Enterprise
For larger orgs with SSO, SCIM, and compliance requirements.
Custom
- SSO, SCIM, on-prem connect, dedicated success engineer
- SOC 2 report, DPA, custom DPIA support
- Per-region data residency
Sovereign
CanadianClaw
For regulated, public-sector, or Canadian-data workloads.
Custom · single-tenant annual
- Single-tenant BC Canada (Cube Global colo)
- Canadian-entity MSA, enforceable in BC courts
- No-logging contractual posture, no US cross-border flows
See full comparison on /pricing. Sovereign option detail on /sovereign.
Two paths. Pick one.
Start free if you want to get your hands dirty. Walkthrough if you want to see it run on your actual use case first.
Start free
One workspace. Three agents. The full skill catalog. Build something this afternoon.
Book a walkthrough
Twenty minutes. We show you your actual use case running in a live workspace.