You have reached this web page by typing "example.com",\n"example.net",\n or "example.org" into your web browser.
\nThese domain names are reserved for use in documentation and are not\navailable\n for registration.
\n\n\n```\n\nNo key, no config. That's the HTML example.com served in 2002. Still not for sale, by the way.\n\nWayback keeps one capture per year by default, so three rows are three years. Its index is in no hurry. Lookups close to a minute happen, so Wayback waits 60 seconds. So do Common Crawl, whose index likes to think it over, and Archive.today. The rest get ten. More on listings and reading: [Snapshots](https://archives.agntn.dev/guide/snapshots), [Reading content](https://archives.agntn.dev/guide/content).\n\n## π What changed?\n\n```ts\nimport { createArchive, diffArchivedContent, providers } from \"@agntn/archives\";\n\nconst archive = createArchive(providers.wayback());\nconst before = await archive.getContent(\"https://example.com/\", { timestamp: \"2020\" });\nconst after = await archive.getContent(\"https://example.com/\", { timestamp: \"2026\" });\n\nconsole.log(diffArchivedContent(before, after).patch);\n```\n\n```\n--- before\t2020-12-31T23:59:42Z\n+++ after\t2026-09-29T05:56:21Z\n@@ -1,8 +1,2 @@\n-Example Domain\n-\n-Example Domain\n-\n-This domain is for use in illustrative examples in documents. You may use this\n-domain in literature without prior coordination or asking for permission.\n-\n-More information...\n\\ No newline at end of file\n+Example Domain This domain is for use in documentation examples without needing permission. This is not a service, avoid relying on it for testing and monitoring purposes.\n+Learn more\n\\ No newline at end of file\n```\n\nEven example.com rewrites its homepage. You asked for years, the header shows the captures you actually got.\n\nPass the full URL here. A bare `example.com` can be `http://example.com:80/` one year and `http://www.example.com/` the next. The diff refuses to compare two different URLs. Want the markup, scripts and comments too? `{ format: \"raw\" }`. That's where the fun stuff hides anyway ;) More: [Comparing captures](https://archives.agntn.dev/guide/diff).\n\n## π§ Library\n\n```ts\nconst archive = createArchive(providers.all());\nconst response = await archive.snapshots(\"example.com\");\n\nresponse.pages; // everything the archives found\nresponse._meta?.errors; // [\"wayback: ...timeout\", ...]\nresponse._meta?.unsupportedProviders; // [{ provider: \"webcite\", reason: \"...\" }]\n```\n\nOn a bad day three archives timed out here and it still came back with 1145 pages. `snapshots()` and `content()` don't throw. Check `success`. `getPages()` and `getContent()` throw instead.\n\nWant a few providers, not all? `createArchive(Promise.all([providers.wayback(), providers.arquivo()]))`. Cache, config files and every option: [Configuration](https://archives.agntn.dev/guide/configuration).\n\n## πΊοΈ Providers\n\n| Provider | Factory | Reads bodies | Needs | In `all()` |\n| -------------------- | -------------------------- | ------------- | ----------------------- | ---------- |\n| Wayback Machine | `providers.wayback()` | yes | nothing | yes |\n| Arquivo.pt | `providers.arquivo()` | yes | nothing | yes |\n| Webarchiv Γsterreich | `providers.webarchiv()` | yes | an exact URL | yes |\n| Archive.today | `providers.archiveToday()` | rendered page | nothing | yes |\n| Common Crawl | `providers.commoncrawl()` | yes | nothing | yes |\n| WebCite | `providers.webcite()` | no | no listing API at all | yes |\n| Archive-It | `providers.archiveIt()` | yes | a `collection` ID | no |\n| Conifer | `providers.conifer()` | no | `user` and `collection` | no |\n| Memento | `providers.memento()` | yes | nothing | no |\n| Perma.cc | `providers.permacc()` | no | an `apiKey` | no |\n\nMemento goes through ODU's MemGator, which already asks several archives. Put it in `all()` and you'd get everything twice. Archive.today has no raw endpoint, so you get the page as it renders it. Running your own MemGator? `providers.memento({ baseUrl })` takes it, over HTTPS unless it's local.\n\nSome of these are history themselves. The original Memento Time Travel is gone. WebCite stopped taking new pages around 2019. Conifer only serves existing collections, read-only. Quirks per archive: [Providers](https://archives.agntn.dev/providers).\n\n## π€ Agents\n\n```bash\narchives mcp\nomp install github:agntn/archives\npi install git:github.com/agntn/archives\n```\n\n```json\n{\n \"mcpServers\": {\n \"archives\": { \"command\": \"npx\", \"args\": [\"-y\", \"@agntn/archives\", \"mcp\"] }\n }\n}\n```\n\nFour read-only tools: snapshots, content, diff and providers. An archived body comes back fenced as untrusted data. It's a recording of a web page, not a message for the model. The Perma.cc key lives in `PERMA_CC_API_KEY`, never in a tool argument.\n\nOne thing to know. `archives mcp` reads its config from your home directory, not from the project your client has open. Browsing someone's repo shouldn't run their `archives.config.ts`. [Agents guide](https://archives.agntn.dev/guide/agents).\n\nThe docs site also has the [Evidence Room](https://archives.agntn.dev/evidence), the same idea over WebMCP in the browser. [WebMCP guide](https://archives.agntn.dev/guide/webmcp).\n\n## π« What this does not do\n\nIt doesn't archive anything. You read what others saved, you can't ask for a new capture. And the live web is [@agntn/web](https://github.com/agntn/web)'s job.\n\n## 𧩠Adding a provider\n\nAn eleventh? Extend `BaseProvider`, answer `snapshots()`, pass it to `createArchive()`. That's it. The full contract: [Custom providers](https://archives.agntn.dev/guide/custom).\n\n## π οΈ Development\n\n```bash\npnpm install\npnpm dev # vp test in watch mode\npnpm lint # builds first, then vp lint and vp fmt --check\npnpm test:types # tsc over the library and both extensions\npnpm test # lint, types, tests with coverage\npnpm build # obuild\npnpm docs # docs site and timeline explorer on :3000\n```\n\n## π Thanks\n\nAnthropic and OpenAI back this package through their open source programs, [Claude for Open Source](https://claude.com/contact-sales/claude-for-oss) and [Codex for Open Source](https://developers.openai.com/community/codex-for-oss). Much appreciated <3\n\n## π License\n\n[MIT](./LICENSE)\n","readmeFilename":"README.md"}