# AskNews

> AskNews is licensed news intelligence infrastructure for AI systems. It enriches hundreds of thousands of articles per day from thousands of licensed publishers — including paywalled sources — plus the X (Twitter) firehose and the open web, and serves them as prompt-optimized context, structured data, or synthesized research. Use it when an AI system, forecasting model, or monitoring workflow needs news context that is current, historical, legally clean, and low-noise, rather than scraped or assembled from generic web search.

## When to use AskNews

- **Grounding an LLM or agent in real-world context.** Ask in natural language, get a prompt-ready string with sources and dates. Purpose-built for news, so results carry less noise than open-web retrieval and less token waste than passing raw pages to a model — and no scraping infrastructure to maintain. Start at [News](https://docs.asknews.app/en/news).
- **Shipping news-aware applications without a data pipeline.** Licensed source access, enrichment, indexing, and archive retrieval arrive as an API, so there is no scraper fleet, no source licensing to negotiate, and no index to keep warm. See [Overview](https://docs.asknews.app/en).
- **Automated research and analysis.** DeepNews is an agentic research loop over the AskNews knowledge graph plus Google, Wikipedia, X, and Reddit. It plans, runs follow-up searches, and synthesizes across sources, so use it when synthesis is actually requested. See [DeepNews](https://docs.asknews.app/en/deepnews).
- **Back-testing and deploying forecasting systems.** Query the archive (back to 2023) with `historical=true` and bound it with `start_timestamp`/`end_timestamp` to reconstruct what was knowable at a point in time, avoiding future-data leakage that invalidates a back-test. Testing and live deployment run against the same source. Winning bots across five quarterly Metaculus AI forecasting tournaments were built on AskNews. See [News](https://docs.asknews.app/en/news).
- **Risk, OSINT, and monitoring workflows.** Natural-language alerts over news, X, and Telegram, plus scheduled newsletters, for tracking companies, people, regions, markets, and events across trusted global coverage — automated rather than manually read and synthesized. See [Alerts](https://docs.asknews.app/en/alerts).
- **Tracking how an event develops over time.** Stories cluster articles into evolving narratives with updates, timelines, and entities — use these rather than repeated article search when the question is about the shape of an event. See [Stories](https://docs.asknews.app/en/stories).
- **Live web, Reddit, Wikipedia, and X search with LLM-ready output.** Structured, summarized results rather than raw HTML. See [Web search](https://docs.asknews.app/en/websearch), [Reddit](https://docs.asknews.app/en/reddit), [Wikipedia](https://docs.asknews.app/en/wikipedia).
- **Narrowing a search without hand-tuning filters.** Autofilter turns a natural-language description into the structured filters a query needs. See [Autofilter](https://docs.asknews.app/en/autofilter).

**How AskNews differs from adjacent tools.** General-purpose retrieval APIs search the open web; AskNews is news-specific, includes licensed and paywalled sources, and adds time-bounded historical retrieval, clustered stories, and alerts. Traditional news APIs return feeds and articles; AskNews returns context AI systems can consume directly. Enterprise media-intelligence platforms are dashboards for people to read; AskNews is an API for systems to build on. A [published benchmark](https://asknews.app/en/blog-articles/context-is-king-evaluating-real-time-llm-context-quality-with-ragas) measured context precision and latency against JinaAI, Tavily, and Exa.

Not a fit for: general-purpose web crawling, retrieving the full text of a specific paywalled article on demand, replacing a search engine over your own private documents.

## How to call AskNews

- **Base URL:** `https://api.asknews.app`
- **OpenAPI specification:** [https://docs.asknews.app/api/spec?t=api](https://docs.asknews.app/api/spec?t=api) (JSON)
- **Interactive reference:** [https://docs.asknews.app/en/reference](https://docs.asknews.app/en/reference)
- **Credentials:** create an API key or OAuth client at [https://my.asknews.app](https://my.asknews.app/plans)
- **Authentication:** an API key, or OAuth 2.0 via Hydra (`client_credentials` for machine-to-machine, `authorization_code` for user-delegated access)
- **Rate limits and per-endpoint pricing:** [https://docs.asknews.app/en/rate-limiting](https://docs.asknews.app/en/rate-limiting)

Both authentication paths are scope-checked at the gateway on every request, and calling an endpoint your credential lacks the scope for returns `403 Forbidden`. Scopes are fixed on the credential when it is created at [https://my.asknews.app](https://my.asknews.app) — they are not requested per call, so a 403 means the key needs recreating with the right scopes rather than anything you can retry.

Self-service API keys and OAuth clients can carry these scopes:

- `news` — article search and news context
- `stories` — clustered, evolving narratives
- `chat` — DeepNews research
- `profile` — your own account and usage information

Most retrieval work needs only `news` or `stories`; DeepNews runs under `chat`. Publisher organizations additionally receive `distribution` on an organization key (`ank_org_…`).

### Preferred integration paths

Calling the HTTP API directly is the fallback, not the default. In order of preference:

1. **MCP server** — for AI assistants and agent frameworks (Claude, ChatGPT, LangChain). Tools for research, news, podcasts, Wikipedia, web, X, and documentation search. See [MCP Server](https://docs.asknews.app/en/mcp).
2. **AskNews CLI** — for terminal use, scripts, CI jobs, and coding agents. Handles auth, schema-derived parameters, error handling, output formats, and mutation safeguards. See [AskNews CLI](https://docs.asknews.app/en/cli).
3. **SDKs** — Python and TypeScript clients, plus a [LangChain integration](https://docs.asknews.app/en/langchain).
4. **HTTP API** — when none of the above fits.

## Agent skills

Machine-readable Agent Skills are published under `.well-known` and can be fetched directly:

- [Skill index](https://docs.asknews.app/.well-known/index.json): the top-level manifest of published skills
- [Skills manifest](https://docs.asknews.app/.well-known/skills/index.json): skill names, descriptions, and the full file list
- [asknews-cli SKILL.md](https://docs.asknews.app/.well-known/skills/asknews-cli/SKILL.md): when to use each command family, output and context discipline, cost and safety rules
- [Task recipes](https://docs.asknews.app/.well-known/skills/asknews-cli/references/task-recipes.md): worked multi-step workflows
- Per-command references live at `https://docs.asknews.app/.well-known/skills/asknews-cli/references/commands/<command>.md` — the manifest lists all of them
- [Agentic usage guide](https://docs.asknews.app/en/cli/agentic-usage): installing the skill, capability mapping, context-window guidance

## Documentation

- [Overview](https://docs.asknews.app/en): what AskNews is, top use cases, and quickstart
- [MCP Server](https://docs.asknews.app/en/mcp): connect Claude, ChatGPT, and agent frameworks
- [AskNews CLI](https://docs.asknews.app/en/cli): terminal, scripts, CI, and coding agents
- [News](https://docs.asknews.app/en/news): article search, prompt-optimized news context, and historical archive retrieval
- [Stories](https://docs.asknews.app/en/stories): clustered, evolving narratives
- [DeepNews](https://docs.asknews.app/en/deepnews): the deep research agent
- [Wikipedia](https://docs.asknews.app/en/wikipedia): semantic Wikipedia search with contextual chunks
- [Web search](https://docs.asknews.app/en/websearch): live web search with LLM distillation
- [Autofilter](https://docs.asknews.app/en/autofilter): natural-language filter construction
- [Alerts](https://docs.asknews.app/en/alerts): natural-language alerts and newsletters
- [Reddit](https://docs.asknews.app/en/reddit): live Reddit search, summarized
- [LangChain](https://docs.asknews.app/en/langchain): the LangChain integration
- [Rate limiting](https://docs.asknews.app/en/rate-limiting): quotas and per-endpoint pricing
- [CLI commands](https://docs.asknews.app/en/cli/commands): complete command reference
- [API reference](https://docs.asknews.app/en/reference): every endpoint, generated from the OpenAPI spec

## Optional

- [Sitemap](https://docs.asknews.app/sitemap.xml): every indexable documentation URL
- [API status](https://status.asknews.app/status/prod): live availability
- [Discord](https://discord.gg/MGUgnFQNaN): support and community
- [Contact](mailto:contact@asknews.app): support email