Back to vendors
A

Algolia

Also known as: algolia.com, Algolia Agent Studio, Algolia MCP Server, Algolia AI Search, Algolia Ask AI

Visit site
Entry priceFree Build plan; Grow and Grow Plus $0/mo base, $0.50 or $1.75 per 1,000 extra searches; Elevate quoted; model costs on your own provider keyFull pricing detail

Algolia is the AI search and retrieval platform used by more than 18,000 businesses, and its Agent Studio builds agents grounded in a team's Algolia indices. Agents search catalogs and content, call client side functions and MCP tools with approval for sensitive actions, remember users across conversations, and run on the team's choice of model from Anthropic, OpenAI, Gemini and others. Guardrails, cost limits, analytics and automated evaluations come built in, and agents ship as chat widgets, search bar AI modes and APIs.

Algolia is the AI search and retrieval platform behind search for more than 18,000 customers in over 150 countries, handling over 1.75 trillion queries a year, among them PetSmart, Leroy Merlin, Arc'teryx, Club Med and DocMorris, while Walgreens doubled its search conversion on Algolia. Agent Studio is its platform for building agents grounded in that search, and it reached general availability in 2026. Orange Group runs its agents in regulated account and billing support across several countries; Orange had Agent Studio live in under two weeks and saw a seven point NPS gain in customer support, on an Algolia search base that lifted order rates by up to 63% when customers searched.

A team defines an agent's role, style and limits, picks its tools and model, and publishes it from the dashboard or API. The built in search tool queries the team's indices, a grouped results tool shortlists by theme, client side functions run inside the team's own app for user data, interface updates or authenticated actions, and MCP servers connect outside services, with control over which tools an agent may use and human approval for sensitive actions. Answers draw on product data, reviews, help content, business rules and merchandising signals, and per user memory keeps facts and episodic records of what the agent observed, did and learned across conversations.

Agents run on Anthropic, OpenAI, Azure OpenAI, Gemini, DeepSeek or any OpenAI compatible provider, chosen per agent on the team's own key. Guardrails screen messages and responses against categories the team defines, cost and usage limits cap spend per agent, analytics track conversations, tokens, searches and feedback, and automated evaluations generate test cases from the index and score relevance, faithfulness and tone before deployment. Agents reach people through InstantSearch chat widgets, an AI mode in the search bar, product page questions and DocSearch Ask AI, and other agents can delegate to them over the Agent2Agent protocol.

Algolia holds SOC 2 Type II, ISO 27001 and 27017 and BSI C5 certification. Its services meet HIPAA, GDPR and CCPA requirements, and enterprise accounts add SAML single sign on and encryption at rest.

Search runs in data centers across Europe, the United Kingdom, the United States, Canada, Brazil, the United Arab Emirates, South Africa, India, Singapore, Hong Kong, Japan and Australia, while model calls stay in the region of the team's OpenAI project or Azure resource.

Pricing follows search usage rather than agents: a free Build plan covers 10,000 searches and 50,000 records a month, Grow and Grow Plus start at $0 a month with overage at $0.50 or $1.75 per 1,000 searches, Elevate is quoted on annual contracts, every agent search counts toward usage, and model costs go to the team's own provider.

Agent Studio suits retailers, telecoms and content sites already on Algolia that want shopping assistants, product questions and support agents grounded in their own catalog and content, with the model, guardrails and spend under their control. It is less suited to back office agents that run on schedules or work through a browser, since its agents start from a person's message or another agent's call.

Vendor details

Canonical URL

https://www.algolia.com

Category

Agent builder

Subcategory

AI search and retrieval platform whose Agent Studio builds agents grounded in a team's Algolia indices, with tools, MCP, per user memory, guardrails, evaluations and a choice of model

Funding status

Venture backed private company headquartered in San Francisco.

Company status

independent

Use cases & customers

Primary use cases

conversational shopping assistants and agentic search barsproduct page questions and guided sellingsupport and knowledge agents over help contentdocumentation assistants with DocSearch Ask AIretrieval and search tools for agents built elsewhere

Target customers

ecommerce and retail teamstelecom and media companiesdeveloper documentation teamsproduct and engineering teams building agents

Deployment options

Algolia hosted SaaSSearch data centers in Europe, the UK, the US, Canada, Brazil, the UAE, South Africa, India, Singapore, Hong Kong, Japan and AustraliaModel calls on the customer's own OpenAI, Azure OpenAI or other provider account

Integrations

Algolia indices with hybrid keyword and vector retrieval; MCP servers as agent tools; client side functions in the team's app; Agent2Agent protocol; Anthropic, OpenAI, Azure OpenAI, Google Gemini, DeepSeek and OpenAI compatible providers such as Groq, Mistral AI, OpenRouter and Together AI; InstantSearch chat widgets for JavaScript and React; DocSearch Ask AI; Algolia MCP Server for ChatGPT, Claude and other clients; commerce connectors including Shopify, Salesforce Commerce Cloud, Adobe Commerce and BigCommerce.

In practice

An outdoor gear retailer adds an AI mode to its existing Algolia search bar. A shopper asks for a light backpack for a three day hike, and the agent searches the catalog with in stock filters, groups options by capacity and remembers next visit that the shopper prefers recycled materials.

A telecom company builds a billing support agent on Agent Studio over its help articles and plan catalog, sets guardrails that block answers on topics outside approved sources, and routes plan changes through an MCP tool that waits for an agent's approval.

A developer tools company switches on DocSearch Ask AI for its documentation, runs automated evaluations generated from its docs index to compare two models on faithfulness, and ships the one that scores better.

Agentic Index coverage score

11.5 / 14 capabilities · 82%

Integrations & Tool Calling Full

Agents call a built in search tool that queries the team's indices with query text, facet and numeric filters set at runtime, a grouped results tool, client side functions that run in the team's own app to read user data, update the interface or take authenticated actions, and MCP servers for outside services such as stock levels. Catalogs flow in through connectors for Shopify, Salesforce Commerce Cloud, Adobe Commerce and BigCommerce.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/toolsread 2026-11-19

Workflow Orchestration Full

Agent Studio runs the whole loop for each request: the model chooses tools from their descriptions, Algolia executes the searches, several tools can be called in parallel in one step, and the answer is written from the results, with completions cached to save tokens. Agents also join multi agent systems over the Agent2Agent protocol, where outside orchestrators discover them through agent cards and delegate tasks to them.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studioread 2026-11-19

Knowledge Grounding & RAG Full

Answers are grounded in live Algolia indices, covering product data, reviews, content, help articles, business rules, merchandising and relevance signals, with hybrid keyword and vector retrieval doing the lookup. Index metadata such as facets and searchable attributes is added to the search tool's description automatically, agent specific search settings sit apart from index settings, and predefined filters can always exclude items such as out of stock products.

Sourcealgolia.com/products/ai/agent-studioread 2026-11-19

Human Oversight & Guardrails Full

MCP tools come with control over which tools each agent may use and a human approval step for sensitive actions. Guardrails, open to Agent Studio customers as a beta, classify user messages and agent responses against categories the team defines on a model it picks, checking the full response with prior messages after generation, and teams set approved sources, tone and response boundaries.

Sourcechangelog.algolia.com/label/78139read 2026-11-19

Security, Identity & Governance Full

Algolia holds SOC 2 Type II, ISO 27001 and 27017 and BSI C5 certification, with an annual SOC 2 audit and outside pen tests twice a year. Its services meet HIPAA, GDPR and CCPA requirements, and enterprise accounts add SAML single sign on and encryption at rest with Algolia Vault. Every account gets two factor sign in and keys limited by access control lists, and Agent Studio adds access controls on who changes an agent.

Sourcealgolia.com/distributed-secure/security-complianceread 2026-11-19

Observability & Auditability Full

Agent Studio analytics track conversation volume, conversation depth, token usage, search activity and user feedback over chosen periods, and every stored conversation can be pulled back through the API with its full message history. While building, the live preview panel shows each tool call with its arguments and results, so a team can see why an agent answered as it did.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/analytics.mdread 2026-11-19

Memory & State Persistence Full

Per user memory keeps two kinds of record across conversations: semantic memories of facts that stay useful, such as a shopper being vegetarian, and episodic memories of what the agent observed, thought, did and learned, which the agent saves and searches through built in memory tools. Memory learns from stored conversations, and a secure user token isolates each person's data in multi user apps.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/memory/overview.mdread 2026-11-19

Deployment & Data Residency Partial

Search runs in data centers across Europe, the United Kingdom, the United States, Canada, Brazil, the United Arab Emirates, South Africa, India, Singapore, Hong Kong, Japan and Australia, and model calls stay in the region of the team's OpenAI project or Azure resource. Agent Studio's own processing of conversations and memory has no region a team can pick, and the platform has no self hosted or private deployment.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/llm-providers.mdread 2026-11-19

Prebuilt Agents / Templates / Packs Partial

Quick Setup offers pre configured templates for shopping agents, and ready experiences cover the agentic search bar, product page questions, guided selling and comparisons, plus Shopping, Category and Comparison Guides and DocSearch Ask AI for documentation sites. The set centers on commerce and documentation search, with no wider library of agents for support, sales or internal jobs.

Sourcealgolia.com/products/ai/agent-studioread 2026-11-19

Triggers & Channel Coverage Partial

Conversations start when a shopper or reader types: in InstantSearch chat widgets for JavaScript and React, an AI mode in the search bar with prompt suggestions, product page questions, DocSearch Ask AI, or a call to the completions endpoint from the team's own app. Other agents can hand work over A2A, but no agent starts on a schedule, a webhook or an event in another system.

Sourcealgolia.com/products/ai/agent-studioread 2026-11-19

Model Flexibility & Routing Full

Each agent runs on the provider and model the team picks: Anthropic, OpenAI, Azure OpenAI with any deployed model, Google Gemini, DeepSeek, or any OpenAI compatible provider such as Groq, Mistral AI, OpenRouter or Together AI, on the team's own key. Changes take effect immediately, guardrails can use a different model, and the LLM leaderboard compares models on relevance, hallucination, speed and cost.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/llm-providers.mdread 2026-11-19

APIs / SDKs / MCP Extensibility Full

The Agent Studio REST API creates and calls agents, streams completions and returns stored conversations, and agents speak the Agent2Agent protocol so outside agents can discover and delegate to them. The Algolia MCP Server, in public beta, lets ChatGPT, Claude and other clients search, analyze and monitor indices, and InstantSearch chat widgets, client side tools and API clients in many languages embed agents in a team's own product.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/conversations.mdread 2026-11-19

Testing, Debugging & Optimization Full

Automated evaluations run before deployment, generating test cases from the team's own search index and scoring answers for relevance, faithfulness and tone, and runs can be compared to measure the effect of a prompt, model or settings change. The live preview shows each tool call while building, and message feedback from users feeds the analytics after launch.

Sourcechangelog.algolia.com/label/81420read 2026-11-19

Browser / Computer-use Not documented

Agents act through search queries, client side functions in the team's own app and MCP tools, all of them defined interfaces. Nothing in Agent Studio drives a browser or a desktop, so an agent cannot fill in a third party web form, click through a supplier portal or work in an application that offers neither an API nor an MCP server for it to call on the shopper's behalf.

Sourcealgolia.com/doc/guides/algolia-ai/agent-studio/how-to/toolsread 2026-11-19

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Free Build plan; Grow and Grow Plus $0/mo base, $0.50 or $1.75 per 1,000 extra searches; Elevate quoted; model costs on your own provider key

Search requests, records, recommendation requests and crawls per 1,000; agent searches count as search requests; LLM tokens billed by the customer's provider

Free tier

Included quota

Build: 10K searches, 50K records, 5K recommendation requests, 5K crawls a month. Grow and Grow Plus: 10K searches, 100K records, 10K recommendation requests, 10K crawls a month

What is public

Algolia's Build plan is free with 10,000 searches and 50,000 records a month. Grow and Grow Plus start at $0 a month and charge $0.50 or $1.75 per 1,000 searches beyond 10,000, with records at $0.40 per 1,000.

Billing mechanics

Agent Studio has no separate price on self serve plans: each search an agent runs counts toward search usage, completions are cached to save tokens, and the model runs on the team's own provider key.

Cost watchouts

Language model tokens are paid to the team's own provider on top of Algolia's search charges.

Variable cost rationale

Every agent conversation can run several searches billed per 1,000, and the model tokens land on the customer's own provider bill, so both lines grow with traffic.

Additional watchouts

A busy agent multiplies search requests, since one conversation can trigger several searches.

Overage / add-ons

Grow $0.50 and Grow Plus $1.75 per 1,000 searches; $0.40 per 1,000 records; $0.60 per 1,000 recommendation requests; $0.80 per 1,000 crawls

Sales call required

Mixed (some tiers require a call)

Free / trial

Free Build plan with 10,000 searches and 50,000 records a month, no card; free GPT-4.1 access for testing agents

Lowest paid plan

Grow, $0 a month base with $0.50 per 1,000 searches beyond 10,000

Commercial notes

Elevate, the enterprise AI search tier, is priced by volume on an annual contract through sales.

Key ambiguities

Whether Agent Studio carries a platform fee of its own on Elevate is not public.

Cancellation / refund

Elevate requires an annual contract.

Missing data

Elevate rates and the typical number of searches per agent conversation are not public.

Agentic Index verified 2026-11-19

Alternatives to Algolia

Other products that do the work Algolia does, and what each one is.

  • DataRobot
  • ChippPlatform for building AI agents
  • JogetOpen source AI powered low code platform
  • LyzrEnterprise platform and control plane
  • RetoolInternal tools platform extended into agent orchestration
  • SimOpen source AI workspace

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.