Solutions

Add an AI assistant to your web app that actually does the work

Most in-app assistants answer questions about your software. Merlean operates it. A user types what they want to achieve, and the work happens inside the product they are already logged into.

The difference between answering and doing

A support chatbot bolted onto a product can tell a user that policies are created under Billing, then Policies, then New. It cannot create the policy. The user still has to go and do it, and the assistant has saved them a search rather than the task.

Merlean closes that gap. When a user says "create a vehicle insurance policy from the latest quote and complete payment", Merlean resolves which quote they mean, calls the endpoints your own frontend would call in the order your application expects, respects the permissions attached to that user's session, and reports what it did.

The distinction matters commercially, not just technically. An assistant that answers questions reduces the time spent looking for a feature. An assistant that executes removes the work.

How it gets installed

Merlean connects to your repository through GitHub or GitLab and analyses the application: its HTTP routes, request and response shapes, form fields and validation rules, permission checks, and the multi-step flows those pieces combine into. That analysis produces a machine-readable model of what the application can do and in what order.

You then drop one script tag into your layout. The widget authenticates as the signed-in user using your existing session or token, so Merlean can never do anything that user could not already do by hand.

No per-feature work

You do not write a tool definition for every action. The analysis pass derives them from the application itself, and re-derives them when the application changes.

Your auth, unchanged

Merlean uses the credentials the browser already holds. There is no second permission system to keep in sync and no service account with elevated rights.

Any stack

Laravel, React, Next.js, Django, Rails, Vue, Node.js, WordPress. Merlean reads applications rather than integrating with frameworks.

What changes for the people using your product

Tasks that took a path through six screens become one sentence.
Features that shipped but were never discovered start getting used, because users describe outcomes rather than hunting for the button that produces them.
New employees are productive without first learning your information architecture.
Support load drops for the largest category of tickets, which is not bugs but "how do I".

Where it fits, and where it does not

Merlean is a strong fit for applications with real depth: multi-step workflows, forms with conditional logic, roles that see different things, processes that span several screens. The more capability hidden behind navigation, the more there is to unlock.

It is a weak fit for a single-screen tool where every action is already one click away. There is no complexity gap to close, and a well-placed button will beat a sentence every time.

Frequently asked questions

Do I have to define every action I want the assistant to perform?

No. Merlean derives the available actions from your application's own code during analysis, and re-derives them when the code changes. You are not maintaining a hand-written tool catalogue.

Can the assistant do things the user is not allowed to do?

No. Merlean acts through your application using the signed-in user's own credentials, so every permission check your application already performs still applies. It has no separate privileged path.

How long does it take to get running?

Connecting a repository and running the first analysis takes minutes. Reviewing what Merlean understood, and correcting anything it got wrong, is the part worth spending time on.

Does it work if my application changes frequently?

Yes. Because the model is derived from source rather than from recorded clicks, a re-analysis after a release picks up new routes and changed forms. Interface changes that would break a UI-scripting tool are not a problem.

See it against your own application

Merlean reads your repository and shows you what it understood before anything is exposed to a user. The fastest way to judge the fit is to point it at a real codebase.

Get in touch Read the docs