# web.pkgforge.dev > Open web access for agents: one host per service, each one HTTP API with its own documents. Every service is free to every caller: no key, no account, no quota. The one exception is the operator's: a service can be switched off (503 `disabled`), and a spend breaker can cap each caller IP for a while (429 `rate-limited`); the status below says when either holds. This document is built on each request from the services' own deployed documents and the health clock's latest reading, and keeps no copy of either. It is Markdown; a browser gets the same text as HTML. Start with a service's quick card, the first section of its `SKILL.md` and `llms.txt` and of its section below: three calls to paste, the failures to handle and whether each is retryable, the headers to read, and the paging loop where it pages. Its `llms-full.txt` ends with the pipeline contract: the stable fields, every failure by retry class, and the same loop. ## Services - [fetch.web.pkgforge.dev](https://fetch.web.pkgforge.dev/llms.txt): A transparent HTTP proxy: GET https://fetch.web.pkgforge.dev/ answers with the origin's status, headers and body, streamed, through Cloudflare's own transports first. - [search.web.pkgforge.dev](https://search.web.pkgforge.dev/llms.txt): A search gateway: one query goes to one upstream provider chosen by tier, mode and health, and the answer is one normalised results[] with provenance. - [read.web.pkgforge.dev](https://read.web.pkgforge.dev/llms.txt): A page's main content as Markdown: GET https://read.web.pkgforge.dev/ answers with the origin's own Markdown, else this service's extraction of its HTML, else a provider's rewrite, each named in provenance. - [research.web.pkgforge.dev](https://research.web.pkgforge.dev/llms.txt): A scholarly works broker: a paper search and a lookup of one work by DOI, arXiv id, PMID, PMCID or OpenAlex id, each answered as one normalised works[] with provenance. - [crawl.web.pkgforge.dev](https://crawl.web.pkgforge.dev/llms.txt): Multi-page crawls as jobs: POST https://crawl.web.pkgforge.dev/crawl submits one to a provider, GET /crawl/ polls it and answers its pages as Markdown. - [archive.web.pkgforge.dev](https://archive.web.pkgforge.dev/llms.txt): A web archive broker: the captures of a URL, one archive a page, with `next` walking on across Cloudflare URL Scanner (its scans of the last 30 days, first inside them and the fallback outside), the Wayback Machine, Arquivo.pt and Common Crawl; the capture closest to an instant, the Wayback Machine's and URL Scanner's compared; one capture's archived body; and a save through Save Page Now. - [forge.web.pkgforge.dev](https://forge.web.pkgforge.dev/llms.txt): Forge reads on this service's own accounts, answered as each forge answers: GitHub's REST API under `/github`, gitlab.com under `/gitlab`, codeberg.org under `/codeberg` (paging on this host, answers stored); GHCR's tags, manifests and files by title under `/ghcr`; repology.org under `/repology`, through fetch.web.pkgforge.dev. - [domain.web.pkgforge.dev](https://domain.web.pkgforge.dev/llms.txt): A domain's facts for any caller, each with its source: its category (adult, gambling, security and the rest of Cloudflare Gateway's ids), its DNS records and mail policy, its registration, its TLS certificate and handshake, the names Certificate Transparency logs hold under it, and all of them in one card. - [ip.web.pkgforge.dev](https://ip.web.pkgforge.dev/llms.txt): The caller's own address and what the edge saw of its connection, and any address's network, registry, cloud, listings, reverse names and reputation, each fact with its source. A route not built yet answers 503. ## Status - [status](https://web.pkgforge.dev/status): every service's state, its providers and the quota left, as Markdown - [status.json](https://web.pkgforge.dev/status.json): the same answer as JSON ## Optional - [llms-full.txt](https://web.pkgforge.dev/llms-full.txt): every service's full documents in one file - [fetch.web.pkgforge.dev SKILL.md](https://fetch.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [search.web.pkgforge.dev SKILL.md](https://search.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [read.web.pkgforge.dev SKILL.md](https://read.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [research.web.pkgforge.dev SKILL.md](https://research.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [crawl.web.pkgforge.dev SKILL.md](https://crawl.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [archive.web.pkgforge.dev SKILL.md](https://archive.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [forge.web.pkgforge.dev SKILL.md](https://forge.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [domain.web.pkgforge.dev SKILL.md](https://domain.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service - [ip.web.pkgforge.dev SKILL.md](https://ip.web.pkgforge.dev/SKILL.md): one file an agent reads to use this service