SAN FRANCISCO, CA — Sept 23, 2025 — Algolia, the retrieval platform orchestrating over 1.75 trillion queries each year, trusted by more than 18,000 businesses, and used by millions of developers around the world, today unveiled Algolia Agent Studio.

Platform, product, and data teams face unrelenting pressure to move beyond prototypes and bring agents into real, business workflows. And yet, they encounter the same persistent obstacles: answers that are irrelevant or stale, or simply wrong. While the potential is clear, agents rely on the information available to them. When that data changes, agents can fracture quickly. Responses with no lineage, trace or control have no way to be understood or trusted. These gaps make agents brittle, unreliable, and incapable of operating at scale.

Bharat Guruprakash, Chief Product Officer, Algolia, noted: “The rise of AI agents introduces something entirely new; a class of users that exist alongside humans. A person might issue one or two queries to complete a task. An agent, on the other hand, can issue many thousands of queries to interpret context, to verify output, to refine results, and to orchestrate actions. This explosion in retrieval demand changes everything. Infrastructure built for human-scale interaction is simply not enough; agents require retrieval that is governed, fast, and able to scale to hundreds and thousands of queries per task, per agent, and per second without breaking.”

Current tools do not solve the problem.
Frameworks can be used by teams to wire components together, but they abandon retrieval quality and place the burden of data on to the team. Vector databases offer embeddings, but no understanding of business rules or relevance. Low-code builders help get prototypes created quickly, but there is no reliability in the answers or decisions made. SaaS copilots impose governance, but only within the narrow confines of their own data silos and never across the breadth of an enterprise’s data. The result: front-end teams are often left to stitch together complex pipelines simply to embed an agent into an application.

Introducing Algolia Agent Studio
Algolia’s Agent Studio resolves these challenges by placing its retrieval engine (comprising fast hybrid, vector search and keyword search, enriched with rules and personalization) at the center of the agent’s workflow. This gives teams retrieval they can trust, clarity to observe and debug, and an elegant path from prototype to production.

Agent Studio equips developers and product teams with:

• Retrieval you can trust: Powerful keyword and vector-based NeuralSearch with rules and personalization to keep agents grounded in enterprise data.

• Design that is model-agnostic: ‘Bring your own’ large language model (BYoLLM). Agent Studio handles retrieval, policy, and runtime while remaining decoupled from model providers.

• Native to Model Context Protocol (MCP): Built to align with the MCP so agents can orchestrate context and tools consistently across stacks.

• Orchestration of tools: Connect APIs and actions, enabling agents to retrieve context, reason, and act within governed workflows.

• Simplicity that drops in: Ready-made React components enable teams to embed agents directly into applications as easily as adding a widget, while always powered by Algolia’s retrieval engine.

• Observability built-in: Traces, evaluation harnesses, and A/B testing provide clear visibility into why an agent responds the way it does.

• Readiness for operations: Real-time indexing, schema flexibility, and audit trails ensure production deployments can thrive in dynamic environments.

Guruprakash added: “Most agent platforms stop at the demo. Agent Studio starts from a different assumption: agents are not just another search box. They trigger on events, chain multiple retrieval steps, and require continuity over time. Memory becomes essential; without it, every interaction starts again. By grounding both retrieval and memory in Algolia’s retrieval infrastructure, agents become accurate, adaptive, and trustworthy in production, where it matters most.”

Agent Studio is a capability applicable to all industries. Early customers in the public beta are already using Agent Studio to:

• Build customer support copilots grounded in knowledge bases, customer relationship management (CRM), and ticketing history applications.

• Deliver in-product SaaS assistants that adapt to user roles, entitlements, and data environments.

• Create smart e-commerce, shopping assistants for both shoppers and backend agents for merchandisers, which combine inventory, pricing, and personalization into conversational shopping journeys.

What’s Next
Agent Studio will add persistent memory, policy-based governance, and production-grade evaluation, enabling agents to stay context-aware across sessions and run safely at enterprise scale.

Availability
Developers and builders can start building now; Agent Studio is available today in public beta, with general availability (GA) planned later this year. 

Algolia to attend AI Engineer conference (Paris) 
Algolia will showcase its Agent Studio at AI Engineer Conference in Paris (23-24 September) at Station F. Visitors can explore live demonstrations of Agent Studio, and discover a range of innovative products. 

About Algolia  
Algolia is the leading AI search and retrieval platform, powering 1.75 trillion searches a year for over 18,000 businesses and millions of developers worldwide. Renowned for its developer-first, API-first design and the fastest AI search technology (combining keyword, vector-based NeuralSearch, rules, and personalization). Algolia is the largest hosted search engine available today. Backed by more than a decade of innovation, Algolia delivers end-to-end AI search capabilities, from powering Retrieval-Augmented Generation (RAG) to enabling LLM-powered applications through its MCP server and new products like AskAI and Agent Studio. With unmatched speed, scalability, and simplicity, Algolia continues to redefine how businesses and developers build intelligent, trustworthy experiences. Learn more at www.algolia.com.