A set of agent skills for fetching knowledge on the public web. Compatible with Claude Code and most harnesses.
/dql Query the Diffbot Knowledge Graph in natural language. Claude constructs and runs the DQL query.
/web-search Search the web via the Diffbot Web Search API. Returns a ranked list of 10 results. Chunks per result included to avoid secondary fetch actions.
/extract Fetch and extract structured JSON or markdown from any URL with the Diffbot Extract API.
/entities Identify and resolve named entities in text using the Diffbot NLP API. Helpful for validating LLM responses or pulling in additional context from Diffbot Knowledge Graph. Entities are linked to KG records in addition to returning confidence, salience, and sentiment scores.
/crawl Crawl a website with the Diffbot Crawl API. Helpful for archival or offline site search.
Python 3.10+diffbot-python— Diffbot Python Library
1. Get a Diffbot API token from https://app.diffbot.com/get-started/
2. Open this project in your harness and run any skill.
That's it. Run any skill again and it's ready.
/dql find large tech companies in Austin, Texas
/dql show me CTOs at public biotech companies
/dql recent negative articles about OpenAI
/dql top cities where data scientists work
/dql software startups in Berlin under 100 employees with a female CEO
Claude constructs the DQL query, executes it against the Diffbot API, and returns formatted results. You can ask for the next page, refine the query, or request a different format.
/web-search AI chip startups 2024
/web-search recent earnings reports Tesla
/web-search latest news on Anthropic
Returns ranked results with relevance scores, URLs, publication dates, and content chunks.
/extract https://example.com/article
/extract https://example.com/product-page
Returns a clean markdown rendering of the page by default. Pass Supports auto-detect or forcing a specific extractor (article, product, image, etc.).
/entities Apple CEO Tim Cook announced record quarterly earnings.
/entities Elon Musk founded Tesla and SpaceX.
Returns a table of entities with type, confidence, salience, sentiment, and Diffbot KG ID. Use -f dql to pipe entity IDs directly into a /dql query.
/crawl https://docs.example.com
/crawl https://example.com --url-process-pattern /blog/ --max-to-process 200
/crawl list jobs
/crawl delete job my-job-name
Streams crawl events as they happen. Jobs can be named, listed, and deleted.
DIFFBOT_API_TOKEN=YOUR_DIFFBOT_TOKEN_HERE
The file lives at ~/.diffbot/credentials on your local machine and is never part of this repository.