Algolia MCP Server – Search and Analytics Tools
Algolia's managed MCP services connect AI agents to Algolia Search, Recommend, and analytics workflows. Use Public MCP for controlled application-scoped access to selected indices and Recommend models, or Productivity MCP for internal, user-scoped search and analytics exploration.
Overview
Algolia provides managed MCP services that let AI assistants call Algolia tools
and retrieve structured data from Algolia in real time. The offering has two
modes: Algolia Public MCP for controlled, application-scoped access to selected
indices and Recommend models, and Algolia Productivity MCP for internal,
user-scoped search and analytics workflows.
What the MCP server enables
Public MCP is designed for production-facing or agent-facing retrieval over a
curated subset of Algolia data. It can expose:
- A search tool for each selected index.
- Facet-value search for supported indices.
- Recommend tools for selected Recommend models.
- ChatGPT-compatible
searchandfetchtools when required by the client.
Productivity MCP is designed for internal work by authenticated Algolia users.
Depending on the user's permissions, it can:
- List indices and search across accessible applications and indices.
- Search for facet values.
- Retrieve recommendations.
- Explore analytics such as top searches, no-result queries, click positions,
no-click rate, top filters, top countries, user counts, and search counts. - Validate or analyze search behavior before changes are made elsewhere.
When to use it
Use Public MCP when an AI assistant should answer with results from specific
Algolia indices, such as product discovery, support search, catalog retrieval,
or customer-facing AI experiences. Use Productivity MCP when an internal agent
needs to investigate search performance, understand query trends, inspect
accessible index data, or answer business-intelligence-style questions about
Algolia usage.
Connection and authentication
Public MCP uses generated URLs in the form
https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp. An Algolia application can
have up to five Public MCP server instances. Each instance can expose up to five
indices and selected Recommend models. Client authentication is set to none
because access is governed by the generated MCP URL and the Search key
associated with the selected configuration.
Productivity MCP uses https://mcp.algolia.com/mcp. It requires OAuth sign-in,
inherits the authenticated user's Algolia permissions, and does not require an
OAuth client ID or client secret in compatible clients.
Key considerations
Public MCP is documented as a beta feature. Requests made through Public MCP
count toward normal Algolia usage, and each tool call typically maps to one API
request. Its access is limited by the configured Search key and exposed indices.
Productivity MCP is documented as read-only and suitable for analysis and
exploration; data changes still happen in the dashboard or APIs. Because
Productivity MCP can access anything the signed-in user can access in Algolia,
treat prompts and outputs as internal data. For result fidelity, instruct the
assistant not to filter, reorder, or summarize returned hits unless explicitly
requested.
Supported Transports
streamable_http
URL: https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp
streamable_http
URL: https://mcp.algolia.com/mcp
Frequently Asked Questions
- When should an AI agent use Algolia Public MCP?
- Use Public MCP when an assistant should search a controlled set of Algolia indices or Recommend models, such as product discovery, support search, catalog retrieval, or customer-facing AI experiences backed by curated Algolia data.
- When should an AI agent use Algolia Productivity MCP?
- Use Productivity MCP for internal, user-scoped workflows such as searching accessible indices, exploring index data, investigating top searches, analyzing no-result queries, reviewing click or no-click metrics, and validating search behavior before making changes elsewhere.
- What can an AI agent access through Algolia MCP?
- Public MCP can expose search, facet-value search, recommendations, and ChatGPT-compatible search and fetch tools for configured indices and models. Productivity MCP can expose index search, recommendations, and analytics tools according to the authenticated user's Algolia permissions.
- How is authentication configured for Algolia MCP?
- Public MCP clients use the generated application-specific MCP URL with client authentication set to none; access is governed by the Search-key permissions behind that server configuration. Productivity MCP uses OAuth sign-in and inherits the signed-in user's Algolia permissions.
- Which transport should be used for Algolia MCP?
- Use Streamable HTTP. Connect to the generated Public MCP URL for application-scoped retrieval, or connect to https://mcp.algolia.com/mcp for user-scoped Productivity MCP. Algolia documents both as managed services, so no local stdio server is required for these offerings.