Introducing Context Highlights

Context Highlights is a new tool designed to help AI agents process web pages more efficiently by extracting query-relevant passages while retaining enough surrounding material for interpretation, reducing context size, cost, and latency in long-running tasks. In tests on Exa’s 250-pair WebCode benchmark, it achieved 76.0% groundedness with an average of 229 output tokens, roughly matching Exa Highlights’ 75.6% while using about 20% fewer tokens and outperforming several alternatives in evidence preservation. On 1,000 SimpleQA Verified questions, Context Highlights produced 76.7% answer accuracy using 183 average tokens, exceeding a Zilliz extractor configuration that used substantially more tokens. The tool is available through Claude Code, Codex, and the Context API, can be paired with Context Search, and is planned for deeper integration into search and research workflows, with future development aimed at incorporating an agent’s broader task context and existing knowledge.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.