RELAYTO developer docs
Every site published on RELAYTO has a public, read-only API, an MCP server and Markdown for AI agents, with no API key. To work in your own workspace, connect an agent to the RELAYTO MCP server.
npx relayto search "open enrollment" --site benefits.relayto.comRead any site published on RELAYTO
Every site on a *.relayto.com address or on its publisher’s own domain answers the same public addresses for AI agents and crawlers. They serve only what any visitor can open without signing in; gated content is left out on purpose.
Replace <site> with the site’s host name, for example benefits.relayto.com.
https://<site>/llms.txt- The site’s hubs, documents and pages, each with a link to its Markdown.
<page URL>.md- Any page as Markdown.
Accept: text/markdown- Any page URL requested with this header answers in Markdown.
https://<site>/.well-known/ai-catalog.json- A catalog of these entry points for AI agents.
https://<site>/sitemap.xml- Every page URL.
curl https://benefits.relayto.com/llms.txt
curl https://benefits.relayto.com/e/what-is-open-enrollment-4uoe8tjxhumq6.md
curl -H "Accept: text/markdown" https://benefits.relayto.com/e/what-is-open-enrollment-4uoe8tjxhumq6The RELAYTO site API
Each site’s REST API searches and reads its pages, hubs and documents, with the text of each as Markdown. It is versioned and read-only, and needs no API key. Guides to every RELAYTO API: RELAYTO API, MCP server & SDK.
Worked example: the API reference and OpenAPI description of benefits.relayto.com.
curl "https://benefits.relayto.com/api/site/v1/search?q=open%20enrollment&limit=1"{
"query": "open enrollment",
"data": [
{
"kind": "content",
"slug": "what-is-open-enrollment-4uoe8tjxhumq6",
"title": "What Is Open Enrollment?",
"summary": "Your yearly window to choose or change your benefits.",
"format": "video",
"url": "https://benefits.relayto.com/e/what-is-open-enrollment-4uoe8tjxhumq6",
"markdownUrl": "https://benefits.relayto.com/e/what-is-open-enrollment-4uoe8tjxhumq6.md",
"apiUrl": "https://benefits.relayto.com/api/site/v1/contents/what-is-open-enrollment-4uoe8tjxhumq6",
"updated": "2026-11-11T09:32:36Z"
}
],
"pagination": { "limit": 1, "hasMore": true, "nextCursor": "eyJvIjoxLCJzIjoiZWNjZDNkOTg0MzU2In0" },
"links": {
"self": "https://benefits.relayto.com/api/site/v1/search?q=open%20enrollment&limit=1",
"next": "https://benefits.relayto.com/api/site/v1/search?q=open%20enrollment&limit=1&cursor=eyJvIjoxLCJzIjoiZWNjZDNkOTg0MzU2In0"
}
}Operations
| Method | Path | Operation | Returns |
|---|---|---|---|
| GET | / | getSite | The site and its API links |
| GET | /search | searchSite | Search the site by keyword |
| GET | /pages | listPages | The site’s own pages |
| GET | /hubs | listHubs | The site’s hubs |
| GET | /hubs/{hubSlug} | getHub | One hub |
| GET | /hubs/{hubSlug}/items | listHubItems | The items in a hub |
| GET | /contents | listContents | The documents in the site’s library |
| GET | /contents/{contentSlug} | getContent | A document or page with its text |
| POST | /exports | createExport | Start an export of several documents |
| GET | /exports/{exportId} | getExport | An export job and its result |
Details
- Base URL
https://<site>/api/site/v1- OpenAPI
https://<site>/openapi.json, OpenAPI 3.1- Reference
https://<site>/developers, with every operation, parameter and error code- Authentication
- None: no API key, no sign-up. Every operation reads; the one POST,
createExport, only starts a job that collects the text of several documents. - Pagination
limit(1 to 50, default 10) and an opaquecursor: passpagination.nextCursoruntilpagination.hasMoreis false.- Errors
- Every 4xx and 5xx answer is an RFC 9457 problem (
application/problem+json) with a stablecodeand ahinton what to do next. - Rate limits
- 120 requests per 60 seconds per client IP. Every answer carries
RateLimitandRateLimit-Policyheaders; a 429 addsRetry-After. - Versioning
- The major version is in the path. Changes within v1 are additive only, and a version that is retired keeps working for at least 6 months after it is deprecated.
- Sandbox
https://<site>/api/site/sandbox/v1answers every operation with fixed sample data in the same shapes, for testing an integration.
{
"type": "https://benefits.relayto.com/developers#error-not_found",
"title": "Not found",
"status": 404,
"detail": "There is no content \"no-such-document\" on this site.",
"instance": "/api/site/v1/contents/no-such-document",
"code": "not_found",
"hint": "Check the slug with searchSite, listContents or listHubs."
}The relayto SDK and CLI
The official TypeScript SDK and CLI for the site API, on npm as relayto. No runtime dependencies; Node.js 20.3 or later. The SDK follows cursors with async iterators, throws typed errors and retries a 429 after Retry-After.
Every command, method and error class: packages/relayto on GitHub.
npx relayto search "open enrollment" --site benefits.relayto.com
npx relayto hubs --site benefits.relayto.com
npx relayto get https://benefits.relayto.com/e/what-is-open-enrollment-4uoe8tjxhumq6
npx relayto site --site benefits.relayto.comnpm install relaytoimport { RelaytoSite } from 'relayto';
const site = new RelaytoSite('benefits.relayto.com');
// One page of search results
const results = await site.search('open enrollment', { limit: 3 });
for (const item of results.data) console.log(item.title, item.url);
// Every hub, following the cursors
for await (const hub of site.allHubs()) console.log(hub.title, hub.url);
// A document with its text as Markdown
const doc = await site.getContent('what-is-open-enrollment-4uoe8tjxhumq6');
console.log(doc.access === 'open' ? doc.markdown : `Description only: ${doc.url}`);RELAYTO MCP servers
Every published site has its own public MCP server for reading it. RELAYTO’s platform server works in your workspace after you sign in.
Every published site
https://<site>/mcpPublic and read-only, no sign-in. Tools:
search_content,list_hubs,get_content. MCP Registry: com.relayto/site-content.Terminal # Claude Code: one site's MCP server claude mcp add --transport http benefits https://benefits.relayto.com/mcp # A client that only starts local (stdio) servers npx -y relayto-mcp --site benefits.relayto.comYour RELAYTO workspace
https://mcp.relayto.com/mcpSign in with your RELAYTO account, and your agent can search, publish, organize, share and measure content with your permissions: 31 tools. MCP Registry: com.relayto/mcp.
Set up the RELAYTO MCP server
Agent skills, rules and plugin
Skills teach an agent the common RELAYTO jobs. In Claude Code, the plugin installs the platform MCP server and the skills together.
Agent rules for coding tools are in the same repository: AGENTS.md, .claude/CLAUDE.md, .cursor/rules/, .windsurf/rules/ and .github/copilot-instructions.md.
npx skills add relayto/relayto-mcp/plugin install relayto --marketplace relayto/relayto-mcp- relayto-getting-started
Connect a client, pick the workspace, fix sign-in and permission errors
- relayto-read-published-site
Read and cite any site published on RELAYTO through its MCP server, API, llms.txt and Markdown, no account needed
- relayto-publish-content
Create content from a PDF, link, HTML or post; edit pages and settings; workspace files; forms
- relayto-hubs-and-access
Build and arrange hubs, set privacy, invite people
- relayto-analytics
Read and report content, hub and workspace analytics
Source, packages and registry
The kit is open source under the MIT license: the SDK and CLI, the MCP bridge, the agent skills, rules and plugin.
- relayto/relayto-mcp
GitHub: source and issues for the SDK, the CLI, the MCP bridge, skills, agent rules and plugin
- relayto
npm: the SDK and CLI for the site API
- relayto-mcp
npm: both MCP servers for clients that only start local servers
- com.relayto/site-content
MCP Registry: every published site’s MCP server
- com.relayto/mcp
MCP Registry: RELAYTO’s platform MCP server
- RELAYTO API, MCP server & SDK
Guides to the RELAYTO APIs: the MCP server, the Site API, the SDK and CLI, and the OAuth content API