A Microsoft Clarity alternative where you do not pay with your customers' data

Sequential has two parts that work together. The judge reads your sessions and says what went wrong. The agent tools give your agent the same data, as tools over MCP. With both, an agent can go from a problem to a measured fix. You decide how much the agent does.

1. 01

   The agent reads the problems

   It reads the cases with list_findings and get_finding, and the sessions with list_sessions and get_session_full. It reads heatmaps as data with list_heatmap_pages and get_heatmap: the elements that people click, for one page and one device class.

2. 02

   The agent gets the fix brief

   get_fix_brief gives one case as Markdown: what people experienced, where, the runtime evidence and how Sequential will check the fix.

3. 03

   The agent links the fix

   Your agent writes the fix in your own tools, with its own access. Then it links the pull request, commit or issue to the case with link_fix.

4. 04

   Your CI records the release

   Record the release with the Sequential CLI (npx @usesequential/cli releases record, with --finding F-12), or let the agent call record_release with an access token that has releases:write. The release names the case that it fixes. A fix link alone does not start a measurement.

5. 05

   The agent checks the result

   get_verification gives the counts before and after the release, with a status, for example Improved or Fixed. Sequential states what changed. It does not state a cause.

6. 06

   The agent checks the judge

   get_calibration shows how often the judge was right, question by question. With review_session, the agent marks a reading of a session right or wrong. With review_finding, it gives a verdict on a case: real, noise, not a problem or a duplicate.

• You choose the agent. The agent, its schedule and what it may do in your own tools are your choice. Sequential has no scheduler.
• You choose the permissions. An owner or admin makes the token and picks its permissions. The path /mcp/readonly does not list the tools that change data, with one exception: getting a fix brief marks the case as handed off.
• Reviews need consent. Through the consent page, the review tools need the box "Also: it can review cases and readings". An access token needs the review permissions. Sequential counts an agent's checks of the judge apart from people's checks.
• A person decides on dismissals. An agent's "noise", "not a problem" and "duplicate" are suggestions. A person clicks Accept or Keep open. An owner or admin can turn on Let agents dismiss cases. It is off by default.
• A person confirms proposals. When an agent asks for a fix draft with request_fix_draft, nothing is sent until an owner or admin approves it. Changes that an agent proposes with propose_changes wait until a person confirms them.

Sequential allows 120 requests a minute for each token or connection, and get_heatmap gives heatmaps as data.

• Ranked sessions. Sequential ranks new sessions. In each judged replay, the judge's reading shows what the person came to do, how the session ended and where the person struggled.
• Cases counted exactly. Sequential groups the same problem across sessions into one case. It counts exactly how many people had the problem. By default, it checks every 15 minutes for exceptions, failed requests, dead clicks, rage clicks and dead ends.
• A check after the fix. Record the release, and name the case that the release fixes. Sequential then counts the same problem before and after that release. The result shows a status, for example Improved or Fixed.
• Data for your coding agent. Connect Claude Code, Cursor, Codex or another MCP client. Your agent gets the problem, how many people had it and the replays, and it gets heatmaps as data.
• Masking on the device. New projects use strict masking by default. Strict masking masks all text, every input and most attribute values while the device records.
• Web and mobile in one player. PostHog's open-source iOS and Android SDKs send events and screen recordings to Sequential. Mobile replays play in the same player as web replays. Each web replay also has DevTools.

There are two ways to switch. In the first way, your coding agent does the setup, and you claim the project. The second way is one script tag that you paste.

1. 01

   Give your agent the prompt

   Copy the prompt below into your coding agent, and put your work email in it.

2. 02

   The agent creates a trial

   The agent creates a trial project for your email address. It needs no account and no token.

3. 03

   The agent adds Sequential

   The agent installs the SDK for your app and checks that events arrive. To remove the Clarity tracking code at the same time, tell your agent.

4. 04

   You claim the trial

   Open the link in the email "A Sequential project was set up for this address". Sign in, or create an account, with the same email address. Enter a Team name, and click Claim the project.

Set up Sequential in this app. Read https://www.usesequential.com/docs/agent-trial.md, create a trial for <your work email>, install the SDK for this app's platform, check that events arrive (ask me to open the app if you cannot), and tell me to open the claim email.

• Before the claim. Websites only on localhost, no mobile recordings, at most 100 sessions, no AI judging.
• Claim in time. Claim the trial within 72 hours, or Sequential deletes it. If you do nothing, a reminder email comes after 48 hours.
• After the claim. You become the owner of its organisation, and recording and AI start to work as in any project. The claim stops the localhost rule and the limit of 100 sessions.
• The terms. If Sequential added the trial to your account when you signed in, accept the terms first. Until you accept the terms of a trial added at sign-in, the app shows none of its data.

1. 01

   Create your account

   Create an account in the hosted preview with your work email and a password. Setup has four steps: Organisation, Project, Install and Agent.

2. 02

   Add the Sequential tag

   Paste the sequential.js tag one time into your site's head, where the Clarity tracking code was. If you use Google Tag Manager, use the tag's Google Tag Manager (Custom HTML tag) version. If your app has a build step, use the npm package @usesequential/browser.

3. 03

   See the first events

   When the first event arrives, the Install step of setup changes from "Waiting for the first event." to "It works". Events usually show within about a minute.

<script async src="https://cdn.usesequential.com/sdk/sequential/v1/sequential.js"
        data-key="YOUR_SEQUENTIAL_INGEST_KEY"
        data-api="https://YOUR_SEQUENTIAL_API_HOST"></script>

Sequential is not affiliated with Microsoft Clarity. Microsoft Clarity is a trademark of its owner. Facts about Microsoft Clarity come from its public pages, checked on 2026-11-07.