Search APIs and models for embedding, reranking, and reading web content in AI applications.
Jina AI provides search-focused models and APIs for professional developers building retrieval and AI applications. Reader converts supplied URLs into text for a downstream model, Search returns web results, Embeddings converts supported text or image inputs into vectors, and Reranker orders candidate documents against a query. Connect through HTTP APIs or the Jina MCP endpoint.
Choose the components your pipeline needs: page preparation, candidate retrieval, semantic matching, or relevance ordering. Reader’s basic URL-reading access can be used without a key; keyed services use token balances and endpoint-specific rate limits. Check the extracted material and ranking results before using them as evidence in an answer. Hosted API access is separate from Elastic Cloud inference and licensed on-premise model deployment.
Fetch a supplied URL and convert its content into text for LLM input. Configure format, token budget, timeouts, and selectors for supported page content. A target website may still block access or return incomplete information.
Retrieve web search results as text suitable for a downstream LLM. Result details can include URLs, titles, content, and available timestamps; verify important findings at the underlying pages.
Choose a supporting embedding model to turn text or image inputs into fixed-length vectors for semantic retrieval and matching. Check the selected model’s supported inputs and task configuration.
Rank a supplied set of candidate documents against a query after initial retrieval. Relevance scores and ordering help select material; they do not verify the documents’ factual accuracy.
Jina provides an MCP server at mcp.jina.ai for connecting its APIs to LLMs. API access is available through documented HTTP endpoints.
Fetch a page with the Reader API and pass its extracted text into a summarization, question-answering, or other LLM workflow.
Gather candidate sources with Search and read the relevant pages before passing material to an answer-generating model. Account for access restrictions, caching, and source reliability rather than assuming search output is complete or current in every case.
Embed text or image data as vectors and use those representations in a retrieval or matching pipeline.
Send a query and candidate documents to the Reranker API to sort results by query relevance after an initial retrieval step.
Configure the Jina MCP server as an MCP endpoint so an LLM workflow can use Jina APIs.
Use Reader at r.jina.ai, Search at s.jina.ai, and the Embeddings and Reranker APIs at api.jina.ai. The Jina MCP endpoint lets an LLM workflow access these services.
Reader converts a supplied URL into text for LLM input. When configured for JSON output, the response includes the URL, title, extracted content, and a timestamp when available.
Basic URL reading through r.jina.ai is available without a key, under its IP-based limit. Search, Embeddings, and Reranker require an API key. Limits differ by endpoint and key tier, and both request and token limits can apply; do not use the Reader allowance as a quota for every API.
Some embedding models accept text and images. Select the model appropriate for your inputs rather than assuming every embedding model is multimodal.
Basic anonymous Reader use is free. Keyed API use is metered in tokens, and paid token top-ups provide continued access after an introductory balance is consumed. The new-key free tokens are a finite grant, not a recurring monthly allowance. Check the endpoint’s accounting rules and rate limits when estimating a workload.
No. Token purchases cover the hosted Jina APIs. Model weights have their own license terms, and commercial on-premise delivery is licensed separately from hosted tokens or Elastic Cloud billing. Choose the deployment path and check the selected model’s terms before self-hosting.
Os dados de tráfego são apenas para referência.
www.voyageai.com
Voyage AI provides embedding models and rerankers for improving search and retrieval in AI applications. It is designed for teams building retrieval-augmented generation and other applications that use unstructured data.
www.ovhcloud.com
OVHcloud AI Endpoints provides serverless inference APIs for a selection of generative AI models, including LLMs, voice, document, and image analysis models. It helps developers add AI capabilities to applications through OpenAI-compatible APIs and supported integrations.
www.mixedbread.com
Mixedbread is a multimodal search API and retrieval platform for applications and AI agents. It processes files such as PDFs, documents, images, code, audio, and video, then returns searchable, ranked evidence for downstream tasks.
