Introducing Highlights: the context that matters

Ground your AI agents. In the live web, not guesses.

Give agents a web search tool and a fetch tool that return clean, relevant content with sources. Fewer hallucinations and fewer wasted tokens, through one SDK or MCP server.

10,000 free search results a month · no card required

Support agenttools: web_search, web_fetch

grounded · 2 sources

What changed in the Acme API v4 release?

web_search“Acme API v4 release changes”3 results · highlights

web_fetchacme.com/changelog/v4markdown · 1.2s

v4 replaces API keys with scoped tokens and adds cursor pagination to every list endpoint [1]. Rate limits now apply per token instead of per account [2].

[1] acme.com/changelog/v4[2] docs.acme.com/rate-limits

web_search · web_fetch

Agents grounded with Context.dev

  • Aldena
  • Scira AI
  • Squad
  • Tinfoil
  • freecode
  • Construct
“This has allowed us to equip the agent with up-to-date context on the hardware, tools, frameworks, etc. it’s using, and thus make it massively more reliable.”
Aarav GuptaFounder and CEO, SpikonadoRead the Spikonado story
lower total LLM and retrieval costs for freecode, fees included
8.9%
for Scira AI to add, test, and deploy web search
< 10 min

Two tools every agent needs.

Search finds the right pages and returns highlights, the passages that match the query. Fetch reads any URL as Markdown. Your agent answers from both and cites them.

  • SearchRanked results with relevance labels
  • HighlightOnly the passages that match the query
  • FetchAny URL as clean Markdown
npm install context.dev

tools.ts

import ContextDev from 'context.dev';

const client = new ContextDev({ apiKey: 'YOUR_API_KEY' });

const search = await client.web.search({
  query: 'What changed in the Acme API v4 release?',
  highlightsOptions: { enabled: true },
});

const sources = search.results.filter((r) => r.relevance === 'high');

const page = await client.web.scrape({
  url: sources[0].url,
  formats: { markdown: true },
});

Agents grounded with Context.dev.

Where our customers give agents the live web, from coding agents to private search.

  • Coding agents

    Give every model the same web_search and web_fetch tools, with tests included.

    Aldena
  • AI search engines

    Real-time web search and retrieval behind an AI answer engine.

    Scira AI
  • Private search agents

    Scraping and metadata with zero data retention for a privacy-first agent.

    Tinfoil
  • Agent cost control

    Highlights cut input tokens 9.8% across replays of real coding sessions.

    freecode
  • Engineering agents

    Current docs on the tools and frameworks an agent is working with.

    Spikonado
  • Consumer assistants

    Ground travel answers in Reddit and Quora discussions, with citations.

    TravelSmart

Skip the tool plumbing you’d have to maintain.

What it takes to give an agent reliable web access yourself, and what Context.dev does instead.

CompareStitched-together toolsbuild and maintainContext.devone API key
SearchA search API plus your own rankingRanked results with relevance labels
Reading pagesA scraper, headless browsers, and proxiesAny URL as Markdown, rendering included
Context sizeFull pages pushed into the promptHighlights with only the matching passages
Where agents runCustom tool wrappers for every frameworkSDKs in five languages and an MCP server
PrivacyData retained by every vendor in the chainZero data retention when you need it

AI grounding, answered.

More details in the API documentation.

What is AI grounding?

AI grounding connects a model’s answer to real, current sources instead of only its training data. Grounded agents search the web, read the relevant pages, and cite them. Context.dev gives agents both tools through one API.

How does grounding reduce AI hallucinations?

A grounded agent answers from pages it just retrieved instead of guessing from training data, and it can cite those pages. Highlights keep only the passages that match the query, so the model reads less noise. freecode measured 9.8% fewer input tokens with Highlights.

How is this different from grounding with Google Search?

Grounding with Google Search is built into Gemini models. Context.dev works with any model or agent framework: call Search and Scrape from your own tool loop or connect the MCP server, and use the pages and highlights however you want.

Can I give my agent web access over MCP?

Yes. Add https://mcp.context.dev/mcp to Claude Code, Cursor, Codex, or another MCP client. Sign-in uses OAuth, so no API key goes into the config file.

How much does grounding cost?

Search costs 1 credit per 10 results, and Highlights add 1 credit per 10 results. Fetching a page as Markdown costs 1 credit. The free plan includes 1,000 credits every month.

Ground your agents today.

1,000 free credits every month, enough for 10,000 search results. No credit card required.