Every team building AI features runs into the same two brick walls:

1. Information retrieval fails when users don’t know the “right” words. You might stuff your index records with keywords, but if the user types “checkout button isn’t clickable” instead of “click analytics configuration”, they’ll never find the solution.
2. Manual data prep slows teams down. Adding attributes, tags, or transformations without AI help usually means hours of poring over spreadsheets or writing custom scripts.

We built Agent Studio to tear down those walls. It’s a new layer on top of Algolia’s structured search indexes that lets you create agents to:

• read directly from the data you already index
• trigger transformations from natural-language instructions
• operate inside production contexts like enrichment pipelines, ecommerce search, and admin dashboards

The flow of Agent Studio

In short: Agent Studio is a new tool to turn retrieval into action, complete with the quality-of-life features you’ve come to expect from Algolia.

To test it out and prove its worth, we do more than ship it out to customers — we run it ourselves. (If you’re not familiar, this is called dogfooding, and it’s a great way for us to be confident that you’ll like using the tool too.) Here are three tools we’ve put into production with Agent Studio.

Ask AI

The problem

Documentation can be a maze. If you don’t know the project’s internal jargon, you’ll spend hours searching or opening support tickets.

The solution

With Agent Studio, we built Ask AI right into DocSearch. It takes a natural-language question, grounds the response in the indexed docs, and returns both a synthesized answer and the supporting excerpt.

DocSearch is now seamlessly integrated with Ask AI, a conversational assistant with full docs knowledge built with Agent Studio

Key takeaways

• Resilient to change: when phrasing shifts, the agent adapts.
• Fast: response times match DocSearch expectations, under 20ms on average.
• Safe: if it can’t generate a real grounded result, the agent falls back to plain search results.

AI Attribute Enrichment

The problem

Enriching data manually is annoying and time-consuming. Adding tags and price tiers and categories (oh my) to every product by hand eats hours, and the fatigue fosters errors.

The solution

We deployed an agent that transforms plain text descriptions into structured enrichment rules. For example, give it the prompt Tag items under $50 as budget. and it’ll give you the option to auto-apply this precise rule on any incoming records:

{
  "condition": "price < 50",
  "attribute": "tier",
  "value": "budget"
}

This lets you skip previously tedious parts of the process:

Key takeaways

• Governable: outputs reviewable diffs that you can dry-run before applying.
• Not just chat: runs as a background automation tool, not only in conversations.
• Explainable: converts natural language into precise JSON that you can verify represents your goals.

Algolia AI Assist

The problem

Our dashboard is powerful but dense. Analytics and configuration windows can be overwhelming at times. Users know they have options, but they don’t always know what the most impactful thing they could do next is.

The solution

Agent Studio powers AI Assist, an agent that reads context from the current view and surfaces actionable suggestions, like “You have no click analytics enabled — here’s how to configure events.” with appropriate links to guides and tutorials. The agent has all sorts of information about what features are actually available to the user, so its suggestions feel less generic and more intentional. It’s also restricted from implementing the suggestions itself using role-based permissions, so we can use cheaper models and trust it not to make a mess.

Key takeaways

• Contextual: creates different suggestions depending on the user’s dataset, current page, and available features.
• High-impact: prioritizes actions correlated with the stated goals, in this case better engagement.
• Controlled: actions in the search index respect role-based permissions.

Next steps

Agent Studio is in beta today and it’s already powering multiple production tools here at Algolia (DocSearch Ask AI, Attribute Enrichment, and AI Assist) along with tons of production tools built by driven devs like you.

Just to recap: what makes these agents so special is their:

• flexibility — they can show up anywhere, in any form, to do anything
• reliability — they’re grounded in real data, so no hallucinations
• resiliency — they fallback on plain results to avoid breakdowns

Ready to try it yourself? Learn more about Agent Studio today and give it a try.