Anthropic Web Search

Kimi Web Search gives Kimi models access to current information from the web, allowing them to search beyond their training data when a question depends on recent events, changing information, or external sources.

Rather than requiring you to build your own search, crawling, and page-cleaning pipeline, Kimi provides web search as a built-in tool. The model can determine when a search is useful, generate the search request, retrieve relevant web content, and use that information to produce a grounded response. Search results are added to the model’s context, allowing Kimi to reason across the retrieved material before answering.

Use Cases

Current market and product research

Fetch Hive agents can use Kimi Web Search to research markets, products, companies, or technologies using information available on the web right now. Agents can then synthesize those findings into reports, briefs, or structured research outputs.

Competitive research and comparisons

Give a Fetch Hive agent a set of competitors, products, or requirements and let Kimi Web Search retrieve current documentation, announcements, pricing information, and other public sources. The agent can compare the findings and turn them into a structured competitive analysis.

Researching recent changes and announcements

Kimi Web Search is useful for questions where static model knowledge may already be outdated. Fetch Hive agents can search for recent launches, company announcements, documentation changes, industry developments, or other time-sensitive information and summarize what has changed.

Why use Kimi Web Search with Fetch Hive?

Kimi Web Search provides Fetch Hive agents with access to current web information, while Fetch Hive turns that research capability into repeatable workflows.

You can run research agents on demand or on a schedule, combine web research with other Fetch Hive tools, route findings into reports or downstream workflows, and inspect detailed traces to understand how an agent reached its result.

Fetch Hive also records the cost of each run, helping you understand how web research contributes to overall AI spend. This can be particularly useful with Kimi Web Search because retrieved search content is added to the model's prompt context and therefore contributes to token usage.

FAQ

Can Fetch Hive agents use Kimi Web Search for current information?

Yes. Fetch Hive agents can use Kimi Web Search to retrieve current information from the web and incorporate that information into their reasoning and final response.

How does Kimi Web Search work?

Kimi exposes web search as a built-in $web_search tool. When the model determines that web information is useful, it can generate a web search tool call. Kimi then performs the search and makes the retrieved content available to the model so it can continue reasoning and generate its answer.

Can I add Kimi Web Search as a workflow step in Fetch Hive?

Kimi Web Search is available for Fetch Hive agents rather than as a direct workflow step. Agents can decide when web research is required while working toward their objective.

Does Fetch Hive show what happened during a Kimi Web Search run?

Fetch Hive provides trace logs for agent runs so you can inspect the actions taken during execution and troubleshoot unexpected or incomplete outputs. The exact search information and source details available depend on what Kimi returns from the web search operation.

Can I track costs for agents using Kimi Web Search?

Yes. Fetch Hive records costs for agent runs so you can review spend per run, workflow, or user.

Kimi Web Search can consume significantly more input tokens than a normal model request because retrieved web content is added to the conversation context. Kimi also charges a separate fee for each web search call, so Fetch Hive's cost tracking can help you understand the real cost of web-enabled research.