LangSearch

LangSearch is a free web search and semantic rerank API platform designed for developers building AI applications that require access to real-world web data. It provides a natural language web search API that searches billions of web documents, combined with a transformer-based semantic reranker that improves result relevance for LLM-powered applications.

How It Works

• LangSearch Web Search API accepts natural language queries and returns structured search results with enhanced details including document summaries, publication dates, and source URLs. It supports custom filtering by freshness, result count, and summary format, making it suitable for research, news aggregation, and knowledge retrieval tasks.
• The Semantic Rerank API performs secondary optimization on initial search results from keyword search (BM25), vector search, or hybrid search. It assigns a rerankScore from 0 to 1 to each candidate document, reflecting the semantic relevance between the query and the document content.
• Both APIs are accessible via simple REST endpoints with Bearer token authentication, returning JSON responses that integrate directly into LLM tool-calling workflows and application backends.

Key Capabilities

• Hybrid search combining keyword and embedding-based retrieval, using the langsearch-ranker model to boost search result accuracy. This approach captures both exact term matches and semantic similarity for comprehensive results.
• Efficient semantic reranking with a model that uses only 80 million parameters yet achieves ranking performance comparable to larger models with 280 million to 560 million parameters. This design delivers faster inference and lower computational costs suitable for production-scale applications.
• Natural language understanding that allows users to query the search engine in plain conversational language rather than constructing complex keyword queries, making it accessible for AI agents and chatbots that process user intent directly.
• Long-text summary generation that condenses raw web content into readable markdown-formatted summaries, providing LLMs with concise structured context rather than raw HTML or lengthy articles.

Integrations

• Direct REST API with JSON request and response formats, compatible with any programming language and HTTP client.
• LLM function call integration optimized for OpenAI and LangChain, enabling AI models to search the web and retrieve real-time information as part of their reasoning process.
• Ready-to-use AI agent plugins for Coze, Dify, Wordware, and GLIF platforms, allowing users to add web search capabilities to their AI workflows without writing integration code.
• Open source SDKs and examples available on GitHub covering common patterns for both web search and rerank API usage.

Use Cases

• AI Agents that need real-time web data access for tasks such as competitive research, news monitoring, fact checking, and data enrichment during automated workflows.
• Retrieval-Augmented Generation (RAG) systems that require grounded, up-to-date context from the web to improve answer accuracy and reduce hallucination in LLM outputs.
• AI Chatbots that benefit from live search results to answer user questions with current information, including recent events and dynamic content that static training data cannot cover.
• AI Search applications that need accurate, semantically ranked results from millions of web sources, combining keyword relevance with deep semantic understanding.
• Research and analysis tools that aggregate web data with filtering by freshness and relevance, supporting systematic information gathering from diverse sources.

Pricing

• Both the Web Search API and Semantic Rerank API are completely free to use with no subscription fees, credit card requirements, or usage quotas applied.
• Web Search API includes free web searches and free summary content generation with markdown formatting support.
• Semantic Rerank API offers free searches for candidate document sets of 1-50 documents and for larger sets of 50-100 documents, with no per-query charges.