Langsearch

LangSearch positions itself as "the world engine for AGI," providing web search and reranking APIs for LLM applications, enabling models to obtain usable context and source links.

Uses

• Connect AI agents to real-time web information so models can use content from the outside world during reasoning.
• Provide sourced context for scenarios such as retrieval-augmented generation (RAG), research assistants, sales assistants, and programming assistants.

Main capabilities

• Web search: supports titles and source links, search summaries, and full web page body text.
• Result and text control: return up to 50 results per request; the text length of each result can be set (default 5,000 characters, adjustable with max_characters).
• Filtering capabilities: supports domain inclusion and exclusion (includeDomains, excludeDomains), as well as date and date range filtering (freshness supports presets, specific dates, or closed-interval date ranges).
• Access methods: provides an API, MCP, and the ability to add the LangSearch Skill to agent workflows.
• Billing display: the API returns input_tokens and output_tokens; the dashboard shows used tokens, consumption percentage, and the time of the next daily reset.

How to use

• After obtaining a LangSearch API key, run the command on the client or save the configuration according to the official documentation.
• Call the Web Search API from the agent's tools and pass the returned context and source URLs to the model.
• It can also be connected via MCP, or the LangSearch Skill can be added to the agent workflow.

Free plan and limits

• The free plan includes all supported search features: titles and source links, search summaries, full web page body text, per-result text length control, domain inclusion and exclusion, up to 50 results per request, and date and date range filtering.
• No credit card required.
• It uses a daily token quota, resetting daily at 00:00 UTC; input and output token prices are currently 0.
• New accounts are rate-limited to 5 RPS, while existing accounts retain their configured request rate; TPM and daily token limits apply, and all Web Search API keys and MCP calls share the same quota.
• Each successful search counts the query as input tokens, and the returned search content counts as output tokens; titles, URLs, and other metadata are not counted. The longer the returned content, the more quota is consumed.

Applicable scenarios

• Find official documentation, migration guides, and release notes for programming agents.
• Turn questions into sourced briefings: retrieve recent announcements, product launches, and industry reports and hand them to a research agent.
• Learn about customers before a call: search company websites, press releases, and public reports to prepare account briefings for sales agents.

Security and enterprise features

• Type II audit: independently audited enterprise-grade security controls.
• Zero data retention: zero data retention for enterprise search.
• Privacy compliance: GDPR-compliant data processing and protection.
• Availability SLA: service-level commitments for production search workloads.

Other information Official text mentions that "1,000,000 AI agents already use it," and provides a comparison of pricing models with Exa, Tavily, and Brave, based on USD usage rates published on September 13, 2026. Products for independent extraction, mapping, research, and generated answers are separate from Web Search; the context included in LangSearch is not an AI-generated answer and is not an independent extraction product. LangSearch usage is subject to a daily token quota.