The AI automation tool that runs your recurring work — unattended.
Most automation does the easy half — when X happens, do exactly Y. Clourou does the half that needs judgment: on the cadence you set, connected to your tools over MCP, it researches, drafts, checks, and reports back in-app and by email. Unattended, on time, every time.
Start free · model included · pay-as-you-go, no subscription
3 changes worth noting today:
- Acme shipped usage-based billing on their pricing page.
- Globex posted a 2.4 changelog — new SSO + audit logs.
- Initech is hiring 4 GTM roles — likely a push upmarket.
Give your agent real tools — connect anything that speaks MCP, not a fixed vendor integration list
The AI automation tool that does the judgment-shaped half of the work.
Schedule anything
Cron-precise cadences in any timezone — hourly digests, daily reports, weekly rollups. Or a one-time task for tomorrow at 9 that runs once and turns itself off. Run any routine on demand too.
Bring your own tools
Connect any MCP server with an API key or OAuth. Tokens are encrypted at rest and auto-refreshed, so unattended runs just keep working.
Remembers every run
Each routine keeps its own memory between runs — what it already told you, your feedback, where it left off — so runs build on each other instead of repeating themselves. Read, edit, or reset it anytime.
Approve before it acts
Flip on approval mode and the agent drafts emails, posts, and updates instead of sending them. Review the pending actions and approve — it executes exactly what you saw.
Fire it from anywhere
Every routine can have its own secure fire URL. POST to it from an alert, a form, CI, or any app — with an optional note that steers that run.
Runs in the cloud
Durable execution on every cadence — no machine to keep on, no cron to babysit. Failed runs retry; nothing silently disappears.
Delivered where you are
Results land live in-app and in your inbox — each with a full tool-call trace and an honest receipt of what the run did: live web lookups, tools used, real cost. If a run can't be confirmed, it says so instead of pretending.
Teach it skills
Install Agent Skills to give your agent specialized know-how — available to every routine.
Files that persist
A sandboxed workspace plus a My Files store that carries across runs, so artifacts build up over time.
Secure by default
AES-256-GCM encryption for every secret, SSRF-guarded connections, and a spend cap you control.
From idea to running routine in three steps.
Describe the task
Tell the agent what to do in plain English — type it yourself, or just ask the in-app assistant and it sets the routine up for you.
Wire it up
Pick a schedule (or a one-time run, or a webhook trigger), attach the tools it needs over MCP, and add any skills. Done in a minute.
Get results
It runs on time, every time — and delivers the output in-app and to your inbox. Ask the assistant about any run to dig in or fix it.
The recurring work you never get to.
Daily competitor digest
Scan rivals' sites and changelogs, summarize what actually changed.
Inbox triage
Sort overnight email, surface what needs a reply, draft the easy ones.
Weekly metrics report
Pull the numbers, write the narrative, email it to the team.
Standup prep
Yesterday's commits plus today's calendar → a tidy status update.
Content radar
Watch your sources and draft the posts that are worth publishing.
Release notes
Turn a batch of merged PRs into customer-ready notes.
No seats, no task caps, no subscription.
No subscription, ever. Every new account starts with $1 in credit to try it — after that it’s pay-as-you-go: top up a prepaid balance and spend it only as your routines run. Turn on auto-recharge to stay topped up automatically.
Start with $1 in credit — then top up as you go
- Pay only for runs that actually happen
- A capable model included — nothing to set up
- Top up any amount from $20 — no subscription
- Auto-recharge keeps your routines from pausing
- In-app + email delivery
Questions, answered.
What exactly is a routine?
A routine is a task you hand to an AI agent plus a schedule for when it should run. Clourou runs the agent on that cadence, gives it the tools you've attached, and records the result.
How is Clourou different from automation tools like Zapier, Make, or n8n?
Those are rule-based builders: you wire up "when this happens, do exactly that," step by step, and they run only what you wired. Clourou is an AI automation tool built around a reasoning agent — you describe the task in plain language and it works out each run itself, doing the fuzzy, judgment-shaped half the rule builders can't. It's MCP-native, so you bring any tool via API key or OAuth rather than picking from a vendor's list; the model is included; and there are no seat fees or task caps — just a spend cap you set, so an unattended agent never runs past a limit you're comfortable with.
Do I need my own AI model or API key?
No. Clourou comes with a fast, capable model managed for you — there's no separate AI subscription, provider account, or API key to wire up. You only pay for the usage your routines actually consume.
What is MCP, and what can I connect?
MCP (the Model Context Protocol) is an open standard for giving agents tools. Connect any MCP server — web search, email, GitHub, Notion, databases and more — with an API key or OAuth.
Are my data and tokens secure?
Every secret is encrypted at rest with AES-256-GCM, outbound connections are SSRF-guarded, OAuth tokens auto-refresh, and a spend cap you control bounds what any account can run.
How does pricing work?
Every new account starts with $1 in free credit. After that it's pure pay-as-you-go: top up a prepaid balance (minimum $20) and spend it only on the usage your routines run — no subscription. Turn on auto-recharge and we'll top you up automatically when your balance runs low.
Can I run a routine manually?
Yes. Every routine has a Run now button, so you can trigger it on demand and watch the result stream in alongside its scheduled runs. You can also give a routine its own secure fire URL and trigger it from any external service with an HTTP POST.
Can I stop the agent from sending things without my sign-off?
Yes — turn on approval mode for any routine. The agent then drafts irreversible actions (emails, posts, record changes) to a pending list instead of executing them; you review the exact actions on the routine page and approve before anything goes out.
Does a routine remember previous runs?
Yes. Each routine keeps a private memory file it reads at the start of every run and updates at the end — so a news digest won't repeat yesterday's headlines, a tracker knows what it already logged, and your feedback sticks. You can view, edit, or reset the memory anytime.
Stop doing it by hand.
Set up your first routine in a minute. Let the agent handle the rest.