Documentation
Last updated: November 9, 2026
Most people never need this page. The dashboard has a guided onboarding flow that walks you through setup in a few clicks. These docs are the reference version, for people who want the detail.
Open guided onboarding in your dashboard →
1. Quick start
Register at /register with an email, password, and company name, then verify your email with the one-time code we send you. Every self-serve account starts immediately on the permanently free Starter plan, no card and no trial clock.
In Settings, connect your S3 bucket: bucket name, optional prefix, log format (API Gateway or ALB), and AWS region, then grant Clew read access to that bucket (see the next section). Once the connection test passes, Clew queues your first scan right away and checks for new logs every 15 minutes after that.
You can upgrade, downgrade, or add blocking integrations at any time from Settings → Plan & Billing.
2. Connecting your log source
Clew reads your logs by assuming an IAM role inside your own AWS account, not by holding a long-lived access key. Your organisation gets its own External ID, shown read-only on Settings → AWS Access, which generates both policies below with your External ID and Clew's AWS account ID already filled in.
Cross-account IAM role
1. Create an IAM role in your AWS account with this trust policy:
{
"Version": "2012-10-17",
"Statement": [{
"Effect": "Allow",
"Principal": { "AWS": "arn:aws:iam::<CLEW_AWS_ACCOUNT_ID>:root" },
"Action": "sts:AssumeRole",
"Condition": { "StringEquals": { "sts:ExternalId": "<YOUR_EXTERNAL_ID>" } }
}]
}2. Attach this permissions policy to the same role. Add the wafv2 statement only if you want blocking, manual or automatic:
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": ["s3:GetObject", "s3:ListBucket"],
"Resource": ["arn:aws:s3:::YOUR-BUCKET-NAME", "arn:aws:s3:::YOUR-BUCKET-NAME/*"]
},
{
"Effect": "Allow",
"Action": ["wafv2:GetIPSet", "wafv2:UpdateIPSet"],
"Resource": "*"
}
]
}3. Paste the role ARN back into Settings. The connection test runs immediately. If you add blocking later, tick the checkbox on that page and update this same role, you never need a second one.
Legacy: bucket policy, no role
Older accounts may still grant access with a bucket policy naming Clew's shared IAM user as principal. This keeps working if you have not yet switched to a role, but the role above is recommended for every new connection. Contact support@clewsec.com if you need the bucket policy JSON.
3. Enabling AWS API Gateway access logs to S3
Select "API Gateway (apigw)" as the log format in Settings. Clew expects a single-line JSON object per request:
{
"requestTime": "...", "ip": "...", "method": "...", "path": "...",
"status": 200, "responseLength": 512, "latencyMs": 45, "userAgent": "..."
}API Gateway delivers access logs to CloudWatch Logs, not directly to S3, so route that log group into your bucket (a subscription filter with a small Kinesis Firehose delivery stream, or a one-off export task for an initial backfill). On first connection Clew checks a sample of your logs against the format you picked and tells you on Settings if it does not match.
4. Enabling AWS ALB access logs to S3
Select "Application Load Balancer (alb)" as the log format in Settings. An Application Load Balancer can log directly to S3, no Firehose needed: in the load balancer's attributes, enable access logs and point them at your bucket and prefix. Clew parses AWS's standard ALB access log format as-is.
5. How scanning works
Clew checks for new logs every 15 minutes. On a brand-new connection it also looks back over your last 7 days of logs so you are not starting from zero. Clew also takes a focused second look at any single IP making an unusually high number of requests within one scan, so a concentrated attacker spread across a wide window is not missed. Your dashboard's "last scanned" time updates every cycle, even when nothing suspicious is found, so you can see Clew is actively working.
Documented limitation: catching a single concentrated attacker needs a meaningful number of requests from that IP within one 15-minute window. A large number of distinct, low-volume IPs, for example a botnet making one attempt each, can evade detection on Starter, Basic and Growth. Pro and Enterprise add distributed attack clustering, which looks across IPs for that kind of spread-out pattern and raises an alert. Clustering is alert only: Clew does not block IPs automatically from a clustering alert.
6. Understanding verdicts
Every detection gets a confidence score from 0 to 1, which maps to a severity band:
| Confidence | Severity |
|---|---|
| 0.80 and above | Critical |
| 0.60 to 0.79 | High |
| 0.40 to 0.59 | Medium |
| Below 0.40 | Low |
Each verdict's detail page shows a score from every detection agent, a short sample of the raw requests involved, and, on the right plan, a written explanation. Threat types include DoS Flood, DDoS, Brute Force, Credential Stuffing, Bot Activity, Scraping, Port Scan, Enumeration, Sequence Abuse, Web Attack, Geo Anomaly, and Unknown Abuse for anything else. An IP you block manually with no prior detection is labelled Manual Block.
The "cost prevented" figure on a verdict is an estimate, not a measured cost, scaled by confidence using the same assumptions published on the pricing calculator on the homepage.
7. The detection agents
Up to six agents look for a different kind of abuse in each batch of traffic, depending on what the batch looks like, plus one passive agent that recognises known bad IPs across your history but does not vote on its own. Clew combines the scores of the agents that ran into one verdict. Some older material rounds this up to "seven agents", that is imprecise: six are active.
| Agent | What it looks for |
|---|---|
VolumeAgent | DoS / traffic floods |
TemporalAgent | Bot-like timing, off-hours activity |
AuthAgent | Brute force, credential stuffing |
PayloadAgent | SQLi, XSS, path traversal |
SequenceAgent | Endpoint enumeration / scanning |
GeoIPAgent | Geographic anomalies |
KnowledgeAgent | Known bad IPs, passive, casts no vote |
IP reputation. When at least one agent finds something in a batch of traffic, Clew checks the main IP in that batch against AbuseIPDB, a public reputation service, even if the batch ends with no verdict. If the IP is listed with a high abuse score, the confidence of the finding rises by 0.08 and the explanation says so, which can lift a borderline finding over the detection line. Reputation never creates a detection on its own, because an agent must have found something first. Only the IP address is sent, private addresses are never looked up, and if the service is unavailable Clew carries on without it.
8. Threat explanations
Growth and above see a written, plain-English explanation of why a verdict was flagged on its detail page, generated from the IP address and aggregated request metrics only, never raw log content or request bodies. If that is ever unavailable, Clew falls back to an automatically generated description so the field is never empty. Starter and Basic see an upgrade prompt in its place.
9. Alerts
Set an alert email and a severity threshold, all threats or high and critical only, in Settings → Alerts. Every delivery, sent or failed, is logged on the Alerts page, with a button to send yourself a test alert. Free Starter accounts do not receive email alerts, that starts at Basic.
10. Blocking IPs
Manual blocking, a button on a verdict or a form to block an IP directly, is available from the Basic plan up. Automatic blocking, where Clew blocks without a click, is available from Growth up. It is off until you turn it on in Settings (after accepting the Blocking Agreement), and it only blocks verdicts at or above your block confidence, 0.75 by default, so some high verdicts are shown but not blocked. Both push to AWS WAF and Cloudflare, whichever you configure in Settings → WAF Configuration and Cloudflare Configuration. Configure either one, both, or neither, if neither is set up Clew never blocks anything, even on a plan that supports it.
Before your first block, your organisation accepts the Blocking Subscription Agreement once, automatically, the first time you check out on any paid plan.
Unblocking is never tier-gated. Even after a downgrade, you can always remove a block you already placed.
11. Team and roles
Every organisation has three roles:
| Action | Owner | Admin | Viewer |
|---|---|---|---|
| View dashboard, verdicts, and IPs | Yes | Yes | Yes |
| Configure S3, WAF, Cloudflare, alerts | Yes | Yes | No |
| Block or unblock an IP | Yes | Yes | No |
| Invite a team member | Yes | Viewer role only | No |
| Change or remove a member | Yes | No | No |
| View or manage billing | Yes | No | No |
Invites to an email outside your organisation's own domain are capped at the viewer role. Manage your team from Settings → Team Members.
12. Plans, limits, and billing
| Plan | Monthly call volume | History retention |
|---|---|---|
| Starter (free) | 2,000,000 | 7 days |
| Basic | 10,000,000 | 30 days |
| Growth | 50,000,000 | 90 days |
| Pro | 200,000,000 | 365 days |
| Enterprise | No fixed cap | Unlimited |
Going over your cap: Clew sends a warning email at 80% and again at 100%. At 100% scanning pauses, and traffic during the pause is not scanned. Scanning resumes on live traffic when your monthly counter resets or you upgrade to a plan with more volume. There is no overage charge, and a volume top-up for the same plan is not offered.
Pausing scanning yourself. An owner or admin can pause scanning from Settings, under Scanning. While paused, Clew reads no new logs, and the dashboard header shows when the pause started. When you resume, logs written during the pause are skipped and scanning continues from new logs. Saved verdicts, blocking rules, alerts for earlier findings, your account and billing are not affected.
How billing works. Clew bills first and you pay after, with no autopay: nothing is ever charged to your card, UPI or bank account unless you start the payment. Your first payment is made when you pick a paid plan. After that, a bill is issued on the 1st of each billing period and is due on the 20th, with a reminder email on the 17th unless you turned routine billing emails off. Prices for INR billing exclude GST; GST tax invoices will be available later. Payment and billing records are kept for 8 years, as the law requires, even after you delete your account.
If a bill is not paid. Your plan stays active until the 1st of the following month, then the account moves to the free Starter plan. The unpaid bill stays open, your history keeps your previous plan's retention window for 30 more days, and IPs you already blocked stay blocked and can still be unblocked.
Restoring your plan. Open Settings → Plan & Billing, choose your plan, and pay all open bills plus the first period of the plan in a single checkout. Any new paid purchase needs every open bill settled first.
Changing plan. Upgrading takes effect immediately and is paid at once, after every open bill is paid: a higher tier costs the price difference and your next bill date does not change. Moving to an annual plan (from monthly, or from a lower annual plan) costs the full annual price and starts a new year now, and the time left on your old plan is forfeited with nothing credited. Downgrading takes effect from your next bill. Cancelling ends your plan at the end of the period you already paid for. Plan changes are not prorated.
Refunds. A full refund is available within 72 hours of your very first payment, self-serve from Settings, one time. Annual plans are non-refundable after that window. See the Refund Policy. Self-serve payment is through Razorpay: customers billed in India pay in INR, everyone else pays in USD. Only the account owner can see and pay bills, and billing emails can also be sent to selected team members marked as billing contacts.
13. Security and data handling
Clew's access to your logs is read-only unless you separately opt in to blocking, which adds a narrowly scoped WAF permission to the same role. Clew stores parsed log metadata, the verdicts it produces, a short raw sample per detection, and IP reputation history, not your full raw log stream, all subject to your plan's retention window above.
Stored credentials and multi-factor secrets are encrypted at rest, passwords are hashed, never stored in plain text.
Request account deletion any time from Settings. Login is blocked immediately, and your data is permanently deleted within 30 days. See the Privacy Policy for the full list of subprocessors.
14. How the cost prevented estimate works
The "cost prevented" figure, shown on your dashboard and on the homepage calculator, is a modeled estimate, not a measured cost. It is meant to give a rough sense of scale, not an exact accounting of money saved.
Each verdict's contribution is scaled by its own confidence score: a low-confidence detection contributes little, a near-certain one contributes close to the full modeled amount for its threat type, rather than either counting fully or not at all.
The underlying per-request cost assumptions are banded into a low and high range rather than a single number, reflecting published industry research on bot traffic and credential-stuffing outcomes, the same bands shown on the homepage cost calculator. Two cost components feed the model: an infrastructure cost for every malicious request, traffic that consumes compute and bandwidth without ever succeeding, and an account-takeover cost, applied only to the credential-stuffing and brute-force share of traffic, using an assumed rate for how often a stuffing attempt actually compromises an account and an assumed average cost per successful takeover.
The exact numbers behind this model, the abuse-rate bands, the takeover success rate, the cost per compromised account, the infra cost per request, are tuning constants, not guarantees, and may be revised as better data becomes available. This page describes the reasoning rather than the literal values, so it stays accurate even when those constants are tuned.
15. FAQ
Is there a free plan?
Yes, every self-serve signup starts on the permanently free Starter plan, no card, no trial, no expiry.
Which log formats are supported?
AWS API Gateway and AWS Application Load Balancer access logs.
Do you store my raw logs?
No, only parsed metadata, detections, and a short sample per detected verdict.
Can I use Clew without blocking?
Yes, leave your WAF and Cloudflare fields empty and Clew will never take a blocking action, even on a plan that supports it.
What happens if I downgrade?
The downgrade takes effect from your next bill, so your current plan stays active until then. You can always unblock an IP you previously blocked, at any tier.
What happens if I do not pay a bill?
Your plan stays active until the 1st of the following month, then the account moves to the free Starter plan. The bill stays open, and you can restore your plan any time by paying the open bills plus a new period.
What if I remove Clew's access on my side?
Your next scan fails with an access error shown on your dashboard, nothing is deleted on Clew's side until you separately request account deletion.