Stop watching recordings. Get a ranked list of what to fix.

From 400 recordings to one ranked list

Checkout went down last week, and nobody knows why. You have 400 recordings and a standup in 40 minutes. You watch six at double speed, find nothing and go in with a guess.

With Sequential, Monday starts with the list. Findings come from problems in the events your app sends, such as a failed request or a rage click. Each finding counts the people it hit.

After a session ends, Sequential sends a short text summary of its events to a third-party service. That service answers three questions: what did the person come to do, did they get there, and how much did it hurt? Sequential keeps each answer as a probability.

Sequential finds, counts and answers. Your agent does the legwork. You make the call.

What you get

• A ranked list of problems. Each finding names one problem, counts the people it hit and links the replays that show it.
• A verdict on each eligible session. It comes 30 to 35 minutes after the session ends.
• The sessions worth a look, ranked. You can also find sessions from a plain description.
• Replay. Your events show on the timeline, next to a developer-tools panel. Funnels, flows and retention are there too.
• Your own agent, connected. In the Agent step, create a token and copy the setup for Claude Code, Cursor, Codex, Claude Desktop or another MCP client. Your agent reads your findings and sessions and does the work in your tools.

Setup has four steps: Organisation, Project, Install and Agent. The guide "Quickstart: your first hour with Sequential" shows each step.

A week at Northlight Outfitters

Example: Northlight Outfitters, a fictional shop.

Overnight, Sequential raises a finding: a failed request on the shipping page. It hit 38 people. In plain words, people get stuck at the shipping address. One replay shows the moment: the form rejects a postcode with a space in it.

The team’s agent reads the finding and files WEB-231 in their tracker. An engineer reviews the change and ships it in release 2026.10.09.

A week later, Sequential reports what it measured. Before the release, the problem hit 41% of sessions on the shipping page. After the release, it hit 12%. This is a measurement, not a verdict on why.

Sequential found it, counted it and measured it. The agent filed it. People approved it.

What we ask

• Use it on real traffic. Findings come from your own users’ sessions. A test site tells you little.
• Mark each call right or wrong. Every ranked session has a thumbs up and a thumbs down. The header keeps score: "Right N of M you checked". After 30 checks, the Calibration tab shows how often the stated odds held on your own sessions. A thumb never changes a verdict. It measures the judge.
• Check within the window. The preview deletes checks after about 15 days. So the 30 checks must come in that time.

How to start

1. 01

   Create an account

   Give your name, your work email and a password of at least 10 characters.

2. 02

   Verify your email

   Open the link in "Verify your email for Sequential" within 24 hours.

3. 03

   Set up and connect

   Set up your organisation and first project. Then install Sequential and connect your agent.

Does a colleague already use Sequential? An Owner or Admin can invite you from Settings, Organisation, Members.

Good to know

• Sessions. Each organisation gets 5,000 sessions a month, shared by all its projects. The preview takes no payment.
• New organisations. In its first 30 days, a new organisation can also record up to 500 sessions a day (UTC).
• At the limit. At the monthly limit, Sequential stops recording new sessions until the first day of the next month (UTC). Data that you sent before the limit stays.
• Need more sessions? Write to support@usesequential.com.
• Session summaries go to a third-party service. Judging sends a short text summary of each session’s events. An owner or admin can turn automatic judging off for each project.
• Retention. The preview keeps 7 days of data by default. Sequential deletes data one week at a time, so some data stays up to 14 days. Deleted data does not come back: the player shows "This recording has expired."
• Automatic judging has a daily allowance. Each project has its own allowance. When it is used, judging stops until midnight UTC. A new organisation has a smaller allowance in its first 30 days.
• Not every session gets a verdict. A session needs at least three meaningful events and must be from the past 7 days. When there is a queue, a verdict can take longer than 35 minutes.
• No accuracy figure. Sequential gives no accuracy figure. Your team’s thumbs measure it on your own sessions. Today, only Worth watching thumbs go into Calibration.
• Invitations last 48 hours. There are no invite links and no SSO.

Questions