# Onto

> Clean Markdown from any URL. MCP first. Two skills, two jobs — do not mix them.

- [Install — docs SKILL.md](https://docs.buildonto.dev/SKILL.md): Copy-paste MCP Connect + one read_url. Claude Code / Codex one-liners, or Cursor Customize → MCPs with https://api.buildonto.dev/mcp. Not the trusted-fetch skill.
- [Trusted fetch — onto-mcp SKILL.md](https://raw.githubusercontent.com/ravixalgorithm/onto-mcp/main/SKILL.md): Grounding skill in the onto-mcp repo. Prefix public URLs with md.buildonto.dev. Not the install prompt.
- [Connect](https://api.buildonto.dev/mcp): Remote MCP URL.
- [Docs](https://docs.buildonto.dev/): Human steps. Quickstart has three routes: the API, your editor, your own site.

> The AI content negotiation layer. Read any page as clean Markdown with an
> accuracy score, or serve Markdown from your own site so agents stop
> parsing your HTML. Same URL, two faithful answers.

Every page below is also available as Markdown: append `.md` to any URL, or
send `Accept: text/markdown`. The full text of all pages is at
https://docs.buildonto.dev/llms-full.txt.

## Start

The fastest path to a first result.

- [Onto Documentation](https://docs.buildonto.dev/index.md): The AI compatibility layer for the web. Read any site cleanly, serve your own site cleanly.
- [Quickstart](https://docs.buildonto.dev/quickstart.md): Pick a route — the API, your editor, or your own site. Each is a few stops and ends with a check that it worked.
- [How it works](https://docs.buildonto.dev/how-it-works.md): One cleaning engine, three trips through it: a read, your build, and an agent's visit to your site.

## Guides

Whole jobs, end to end. Each one chains several endpoints.

- [Give an agent the web](https://docs.buildonto.dev/guides/agent-web.md): Give an agent clean pages and a trust signal in one tool — and what to do when a page scores badly.
- [Index a whole documentation site](https://docs.buildonto.dev/guides/knowledge-base.md): Every page of a docs site as clean text without writing a crawler: map, batch in fifties, re-read what got cut.
- [Audit a site for agent readability](https://docs.buildonto.dev/guides/audit-a-site.md): Score every page for free, rank the worst, and fix the faults in the order that moves the most pages.

## Read API

Call the API directly over HTTP.

- [Authentication](https://docs.buildonto.dev/api/auth.md): Every Read API request needs a Bearer token in the Authorization header.
- [POST /v1/read](https://docs.buildonto.dev/api/read.md): Fetch any public URL and return clean Markdown plus extraction stats.
- [POST /v1/score](https://docs.buildonto.dev/api/score.md): Score any URL for AI readability — get penalties, benefits, and ranked recommendations.
- [POST /v1/read-and-score](https://docs.buildonto.dev/api/read-and-score.md): Markdown + AIO score + hallucination risk in one request.
- [POST /v1/batch](https://docs.buildonto.dev/api/batch.md): Read, score or extract up to 50 URLs — or a whole site — in one call, at one credit per URL.
- [POST /v1/map](https://docs.buildonto.dev/api/map.md): Discover a site's URLs — fast and cheap, without reading any pages.
- [POST /v1/extract](https://docs.buildonto.dev/api/extract.md): Return the structured data a page already declares — JSON-LD, OpenGraph, and meta tags.
- [GET /v1/usage](https://docs.buildonto.dev/api/usage.md): Inspect your current plan, monthly quota, credit balance, and concurrency state.
- [Error codes](https://docs.buildonto.dev/api/errors.md): Every error returns JSON with a stable code and a human-readable message.
- [Rate limits + credits](https://docs.buildonto.dev/api/limits.md): Three layers — monthly quota, concurrent cap, and credit overflow.
- [Interactive reference](https://docs.buildonto.dev/reference.md): 

## MCP server

Use Onto from an AI coding client.

- [MCP server](https://docs.buildonto.dev/mcp.md): Give any MCP client six Onto tools — by URL with an OAuth sign-in, or locally with npx.
- [Integrations](https://docs.buildonto.dev/integrations.md): Pick your client for its exact install — Claude Code, Claude Desktop, Cursor, Codex, ChatGPT, Windsurf, Cline, Continue, or any MCP client.
- [Set up your client](https://docs.buildonto.dev/mcp/setup.md): Pick your client, install, and check it worked. Most clients only need the URL and a sign-in.
- [Remote connector (OAuth)](https://docs.buildonto.dev/mcp/remote.md): One URL and a sign-in with your Onto account — no key to paste, nothing to install.
- [Tools reference](https://docs.buildonto.dev/mcp/tools.md): The six tools — inputs, what each returns, and what each costs.
- [Configuration](https://docs.buildonto.dev/mcp/configuration.md): The local server's settings, version pinning, and what its errors mean.

## Serve SDK

Make your own site agent-readable.

- [Installation](https://docs.buildonto.dev/sdk/installation.md): Install @ontosdk/next: scaffold, add your site key and the build step, and check agents get Markdown.
- [Configuration](https://docs.buildonto.dev/sdk/configuration.md): onto.config.ts — every field, the llms.txt it generates, and the JSON-LD page types.
- [OntoProvider](https://docs.buildonto.dev/sdk/provider.md): OntoProvider and OntoHead: the JSON-LD and Markdown discovery links they add to your pages.
- [Edge middleware](https://docs.buildonto.dev/sdk/middleware.md): Decides, per request, who gets Markdown — and what goes back, headers and all.
- [Build CLI (onto-next)](https://docs.buildonto.dev/sdk/extractor.md): What onto-next does after next build: which pages it reads, what each .md holds, and what its output means.

## Dashboard

Keys, usage, and agent analytics.

- [Dashboard overview](https://docs.buildonto.dev/control-plane/dashboard.md): A tour of app.buildonto.dev — keys and usage for Read, setup and analytics for Serve.
- [Serve analytics](https://docs.buildonto.dev/control-plane/analytics.md): Which agents read which pages of your site — what's recorded, what's shown, how long it's kept.
- [Context injection](https://docs.buildonto.dev/control-plane/injection.md): Add context to a route's Markdown for agents, without touching the page or redeploying.

## Concepts

The ideas the product is built on.

- [AIO Score](https://docs.buildonto.dev/concepts/aio-score.md): How reliably an agent can read a page, 0 to 100 — every penalty, the grades, and hallucination risk.
- [React Tax](https://docs.buildonto.dev/concepts/react-tax.md): How much of a page is framework rather than content — measured live — and what Onto removes.
- [Content negotiation](https://docs.buildonto.dev/concepts/negotiation.md): One URL, two answers: how a request gets Markdown or HTML, on your site and on these docs.

## API

- Base URL: https://api.buildonto.dev
- Auth: `Authorization: Bearer onto_sk_...`
- OpenAPI + interactive reference: https://docs.buildonto.dev/reference
- MCP endpoint: https://api.buildonto.dev/mcp