canonical: https://jentic.com/apis/marqo.ai/marqo # Marqo Cloud API The Marqo Cloud API creates and manages vector search indexes, adds or replaces documents in them, and runs tensor search that ranks documents by semantic similarity rather than keyword overlap. It lists your indexes, fetches or deletes an individual document by id, and returns recommendations based on documents already in an index. Marqo handles embedding generation internally, so you send text or structured documents and query with natural-language text. ## For AI agents Create and delete vector search indexes, add and replace documents, run tensor search over them by semantic similarity, fetch or delete documents by id, and get recommendations. Marqo generates embeddings internally, so you send and query with plain text. ## Scope Provides vector indexing, tensor search, and recommendations; does not host a chat or language-model completion endpoint or manage billing through this surface. Use for semantic search and retrieval. ## Capabilities - Create a new vector search index with a chosen configuration - Add or replace documents in an index for semantic retrieval - Run tensor search that ranks documents by semantic similarity to a query - List all indexes in your Marqo account - Fetch or delete an individual document by its id - Get recommendations based on documents already in an index ## Use cases ### Semantic search backend for an AI agent Give an AI agent a retrieval layer it can populate and query without hand-wiring the Marqo x-api-key header. Through Jentic the agent discovers the index, document, and search operations by intent and calls them with a validated input schema, so a retrieval-augmented assistant can index a corpus and then run tensor search to fetch the most relevant passages for a question. Example prompt: Create an index, add a batch of support articles as documents, then run a tensor search for 'how do I reset my password' and return the top three matches ### Retrieval-augmented generation Index a knowledge base as documents and run tensor search to retrieve the passages most relevant to a user question, then feed those passages to a language model as grounding context. Because Marqo generates embeddings internally, you index and query with plain text and never manage an embedding pipeline yourself. Example prompt: Search the product-docs index for passages relevant to a customer question and return the top five with their scores ### Index and document management Create and delete indexes, add or replace documents in bulk, and fetch or delete an individual document by id. This lets a content or data pipeline keep a search index in sync with a source of truth, adding new records and removing stale ones as they change. Example prompt: Replace an outdated document in the catalog index with its updated version and confirm the change ### Similar-item recommendations Return recommendations based on documents already in an index, surfacing items semantically similar to a reference set. This backs 'more like this' features, related-content panels, and catalog recommendations that rank by meaning rather than shared tags. Example prompt: Given three product documents a user viewed, return recommendations for similar products from the catalog index ## Key endpoints | Method | Path | Description | | --- | --- | --- | | POST | `/indexes/{indexName}` | Create a new index | | GET | `/indexes` | List all indexes | | DELETE | `/indexes/{indexName}` | Delete an index | | POST | `/indexes/{indexName}/documents` | Add or replace documents in an index | | POST | `/indexes/{indexName}/search` | Run tensor search over an index | | POST | `/indexes/{indexName}/recommend` | Get recommendations based on documents | ## Key resources - **Indexes** — Create, list, and delete vector search indexes - **Documents** — Add, replace, fetch, and delete documents in an index - **Search** — Tensor search that ranks documents by semantic similarity - **Recommendations** — Recommend documents similar to a reference set ## Why Jentic - **Setup:** Wiring the Marqo Cloud API by hand means setting the `x-api-key` header on every request and coordinating index creation, document upload, and search calls in sequence. Through Jentic you install once, import Marqo from the API Directory, store the key once, and your agent calls the index, document, and search operations directly. - **Permission scoping:** You choose the operations your agent may call, so you can allow searching and reading documents while withholding index and document deletion. That way a retrieval assistant can query your data without the ability to destroy an index. - **Credential handling:** Your Marqo Cloud API key is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context. - **Discovery method:** Agents search Jentic by intent such as 'run a semantic search' or 'add documents to an index', and Jentic returns the matching Marqo operation with its input schema so the agent calls the right endpoint without reading the reference docs. ## Related APIs - **Pinecone** — Managed vector database where you supply your own embeddings. - **Meilisearch** — Fast keyword-first search engine with optional semantic search. - **Algolia Search API** — Hosted keyword search tuned for instant, typo-tolerant results. ## FAQ ### What authentication does the Marqo Cloud API use? The Marqo Cloud API authenticates with an API key sent in the `x-api-key` request header, per its OpenAPI spec. You obtain a key from your Marqo Cloud account. Through Jentic the key is stored once and injected at call time, so your agent never places it in the header itself. ### Can I build semantic search with the Marqo Cloud API? Yes. You create an index, add documents to it, and run tensor search that ranks results by semantic similarity to a query. Marqo generates the vector embeddings internally, so you send plain text or structured documents and query with natural-language text rather than managing your own embedding pipeline. ### Is there a Marqo MCP server? You do not need an MCP server to give your agent the Marqo Cloud API. Jentic connects it directly from the API Directory: import it, store your API key once, and your agent calls the index, document, and search operations on demand. Nothing extra runs, and no server tool definitions sit in the agent's context between calls. ### Can I limit what my agent is allowed to do with the Marqo Cloud API? Yes. Write a rule that allows only the operations you need, such as searching an index while leaving out index deletion, so the agent can retrieve documents but cannot destroy your data. Every call it makes is logged, so you can audit exactly which operations ran. ### What are the rate limits for the Marqo Cloud API? The OpenAPI spec does not specify rate limits. Limits depend on your Marqo Cloud plan and the size of your indexes, so check the Marqo documentation at https://docs.marqo.ai/ for the quota that applies to your account. ### How do I run a semantic search with Marqo through Jentic? Search Jentic for 'run a tensor search' to find the search operation, which ranks documents in an index by similarity to your query, and pair it with the add-documents operation to populate the index first. Jentic returns each operation with its input schema so your agent calls it correctly. To run it on your own infrastructure, install Jentic One from its GitHub repo.