Agent Skills
▶ github.com/diffbot/diffbot-skills
Plug Diffbot into your favorite agent harness. Skills abstract and in some cases extend Diffbot APIs for LLM powered web research tasks ranging from web fetching (extract) to web search and entity resolution (NLP).
Compatible Harnesses
- Claude Code
- Github Copilot
- Snowflake Cortex
- Factory.ai (Droid)
- ForgeCode
- pi.dev
- Universal (via npx skills)
Installation
Agent development move fast. Check out the README for the latest installation instructions.
Skills
Give your agent access to Diffbot's Knowledge Graph and web structuring APIs. Query over a trillion facts on organizations, people, and news; Crawl and extract sites to build your own structured knowledge graphs.
Knowledge Graph
/diffbot-news — News and articles
type:Article search, newest-first by default. Filter by mentioned entity, topic taxonomy, publisher, language, date range, or sentiment.
/diffbot-news recent AI news
/diffbot-news negative coverage of OpenAI this month
/diffbot-news what has Sam Altman been quoted saying
/diffbot-news European coverage of the Nvidia earnings/diffbot-organizations — Company search
type:Organization search by industry, headquarters, headcount, revenue, funding, ownership, or leadership.
/diffbot-organizations public semiconductor companies in the US
/diffbot-organizations AI software startups in Berlin under 100 employees
/diffbot-organizations companies like OpenAI, US only
/diffbot-organizations companies Sequoia has invested in
/diffbot-organizations everything Microsoft has acquired/diffbot-people — People search
type:Person search by job title, employer, employer industry, skills, education, location, or nationality. Covers people with a public online professional presence — it is not a people-finder for private individuals.
/diffbot-people who runs Nvidia
/diffbot-people CTOs at biotech companies
/diffbot-people Stanford CS alumni
/diffbot-people female CEOs/diffbot-places — Geographic entities
Cities, counties, states, countries, regions, and points of interest — by name, population, prominence, containing place, or proximity.
/diffbot-places list all countries in Europe
/diffbot-places largest cities in Japan by population
/diffbot-places counties in California
/diffbot-places national parks near Yosemite/diffbot-deals — Funding and M&A
Funding rounds, investments, and acquisitions by industry, date, size, series, investor, or company.
/diffbot-deals AI rounds over $50M this year
/diffbot-deals Sequoia's recent investments
/diffbot-deals largest acquisitions of 2026
/diffbot-deals OpenAI funding history/diffbot-dql — Raw Knowledge Graph queries
Explore all other standard entities in the Diffbot Knowledge Graph not already covered above. There are tons more like CreativeWork, Brand, Patent, SaaS, Technology, even Research. The agent translates your request into DQL, explores the ontology, probes query variants, and exports typed JSON or CSV.
/diffbot-dql show me all brand marks owned by Disney
/diffbot-dql latest published research in artificial intelligenceWeb Structuring
/diffbot-web-search — Live web results
Ranked results with relevance scores, URLs, dates, and content snippets. For general web pages and lookups that aren't news — use /diffbot-news for news.
/diffbot-web-search AI chip startups 2024
/diffbot-web-search recent earnings reports Tesla/diffbot-extract — Structured page content
Fetch and extract structured content from any URL — markdown by default, full JSON on request.
/diffbot-extract https://example.com/article
/diffbot-extract https://example.com/product-page/diffbot-entities — Named entity resolution
Identify and link entities in text to Diffbot KG records. Returns confidence, salience, sentiment, and Diffbot IDs usable in DQL.
/diffbot-entities Apple CEO Tim Cook announced record quarterly earnings.
/diffbot-entities Elon Musk founded Tesla and SpaceX./diffbot-crawl — Site crawling
Crawl a website for structured content and manage crawler jobs.
/diffbot-crawl https://docs.example.com
/diffbot-crawl https://example.com --url-process-pattern /blog/ --max-to-process 200