An agent that acts inside your product. In three days.
An agent in your staging environment, connected to your own front end. It reads your pages, navigates your routes, and runs two actions you choose on day 1.
At the end of day 3, you have
Software running in your environment, and the code in your repository.
An agent running in your staging
Connected to your own front end: React, Vue, Angular or plain JavaScript. It navigates your routes and reads the state of each page.
Two real actions
Actions you choose on day 1, executed inside your product, with user validation before every write. Your permissions are never widened.
A measured baseline
We run your real user questions against the agent and record what it answers, where it stalls, and which documentation gaps cause the stalls.
The code, in your repo
Everything we write lands in your repository. It is yours whether you continue with us or not, and your team can extend it.
A production costing
What it would take to go live: scope, effort and price, handed to you on day 3 with no obligation to proceed.
Your documentation, graded
From the preparation week: which of your documents an AI reads badly, which topics contradict each other, and what is missing.
One preparation week. Then three days with you.
The ingestion and the analysis happen before we arrive, so your three days go into building.
We prepare without you
- We ingest your documentation
- We analyse its quality and find the gaps
- We prepare a set of your real user questions
- You give us read access. Nothing else.
The agent sees
- We show you what we found in your documentation
- We pick the two actions together, and lock the scope
- SDK integration into your front end
- The agent reads your page state and navigates your routes
The agent acts
- The two actions, live in your staging environment
- User validation before every write
- Your JWT is forwarded, your permissions stay yours
- A third action if there is time
Measure and hand over
- Your real questions run against the agent
- A measured baseline: what works, what is missing
- One hour debrief with your decision makers
- What production would take, costed
Your internal cost: about one developer day
There is no requirements document to write first.
What we need from you
- One front end developer, about one day across the three
- Access to a staging or test environment
- Read access to your documentation, for the preparation week
- One decision maker, one hour, at the day 3 debrief
What we bring
- Two AI engineers who have run agents in production for eighteen months
- The AGO platform, the SDK and the LLM, all included
- The preparation week, done before we arrive
- On site or remote, whichever suits you
If the agent does not run the two actions, we do not invoice
The two actions are agreed and locked at the end of day 1, and the scope does not move after that. If they are not working at the end of day 3, you owe us nothing.
We have been running agents in production for eighteen months
The Sprint is the first three days of the method we already run for customers in production.
Mirakl runs an enterprise marketplace platform for customers like Decathlon and Leroy Merlin. We did not start at these numbers. It took eighteen months of continuous work, and the Sprint is the first three days of it.
The Sprint is not for everyone
Three days is short, so it only works when there is documentation to connect and a developer available.
A good fit if
- Your product is complex enough that new users need training
- You have product documentation, even imperfect
- You can free one developer for about a day
- You already have an agent, or a committed plan to build one
Not yet if
- –You have no product documentation at all. There is nothing to connect.
- –You cannot free a developer that week. The Sprint does not work without you.
- –You are looking for a support chatbot beside your product rather than inside it.
- –You are still exploring. Start with a documentation diagnostic instead.
Before you book
The things people ask us on the first call.
What if we want more than two actions?+
Two is what fits in three days without cutting corners. A third one lands if there is time. Everything else is scoped and costed for you on day 3, with no obligation.
Who writes the code?+
We do, with your developer alongside. At the end it sits in your repository and your team knows how to extend it. The code is yours whether you continue with us or not.
Which LLM do you use?+
Whichever you want. We operate it and it is included. If you have a data residency or model governance constraint, we connect to your own endpoint, Bedrock EU or Azure.
Does our data leave Europe?+
No. European hosting, GDPR by design. During the Sprint we work in your staging environment, never in production.
Can we run it on production?+
No. The Sprint runs on your staging or test environment. Going to production is a separate, deliberate step that we cost for you on day 3.
Which front end frameworks do you support?+
React, Vue, Angular, and plain JavaScript or TypeScript. If your application renders in a browser, we can connect to it.
How long between booking and the Sprint?+
Two to four weeks depending on availability. The preparation week runs before your three days and costs you nothing but read access.
Ready to reimagine your customer experience?
Deploy AI agents that truly understand your customers and your business. Book a demo with our team to see Ago in action.