An AI SEO API for agents — MCP today, REST coming soon

Where the AI SEO API stands today

Most "AI SEO API" pages quietly imply a REST endpoint that either doesn't exist or sits behind a sales call. SourceWatch is built the other way around: the access that works right now is the MCP server, and the REST API is on a published roadmap you can hold us to. Here is exactly what is shipped, what is next, and what is planned — in plain terms.

For agents, the MCP server is the API

It's tempting to read "MCP, not REST" as a compromise. It isn't. The Model Context Protocol is an open standard, launched by Anthropic in November 2024 as a way to build secure, two-way connections between data sources and AI-powered tools. It replaces a pile of bespoke one-off integrations with a single protocol — the same role REST played for the last era of the web, but built for agents.

And it isn't a single-vendor bet. Through 2025 the rest of the industry converged on it:

• **OpenAI adopted MCP** in March 2025, integrating it across products including the ChatGPT desktop app.
• **Google DeepMind confirmed Gemini support** in April 2025.
• **Microsoft and major coding platforms** (including Replit and Sourcegraph) followed, wiring MCP into developer tooling.
• The protocol now ships official SDKs in Python, TypeScript, C#, Java and more — so an MCP server is a first-class integration target, not a niche one.

There's a second, quieter advantage. A REST client *reads* data and then a human decides what to do. An MCP client reads the same data and can **act on it in the same loop** — pull your citation gaps and the real queries the models ran, then draft answer-first content against them in Claude Code without leaving the editor. That act-in-the-loop workflow is the second moat, and almost no legacy SEO tool offers it; where a comparable agent layer exists, it's enterprise-only and gated behind a separate subscription. SourceWatch puts it on a self-serve plan.

What the API exposes (today via MCP)

Whether you reach it through the MCP server now or the REST read API later, the underlying data is the same — the two sides of AI search SourceWatch measures. Today, from inside an MCP client, your agent can read:

Visibility & citations

• **Mention & self-citation rate** — how often ChatGPT, Perplexity, Gemini and Claude name you, and whether they cite your own pages.
• **Share of voice** — your slice of brand mentions versus the competitors the models name instead, on a fixed prompt set, over time.
• **The real queries the models ran** — the actual prompts behind each answer, plus sentiment, most-cited domains, and the citation gaps to close.

First-party AI traffic

• **AI-crawler analytics** — which AI bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended and more) read which pages, how often — see AI crawlers.
• **AI-referral visitors** — the real people who arrived from an AI answer, captured first-party instead of lost to "direct." See AI traffic analytics.
• **Verification metadata** — every crawler and visitor checked against the vendor's published IP ranges, so spoofed user-agents don't pollute what your agent reads.

The REST API: what's coming, and what to expect

The public REST read API is the next phase, and it exists for the integrations MCP doesn't naturally cover: a language-agnostic HTTP surface that drops into an existing CI pipeline, a BI tool, or a customer-facing dashboard — anywhere a plain authenticated GET is the right shape. It will be token-authed and account-scoped, the same per-account isolation model the platform already runs on.

What to expect, stated plainly:

• **It is read-first.** Visibility, citations and prompt data out; not a write API.
• **It is not live yet.** Access today is the MCP server. There is no documented REST endpoint to point at right now — and we won't pretend there is.
• **Webhooks follow it.** Push on citation/visibility changes is planned *after* the read API, so you can stop polling once both land.
• **You join a waitlist, not a checkout.** Tell us your stack via contact and we'll bring you in as REST opens up.

Why an AI SEO API matters now

Programmatic access only matters if the channel it measures is real — and AI search has crossed that line. ChatGPT reached more than 800 million weekly active users, announced by OpenAI's Sam Altman at its October 2025 Dev Day — up from 500 million at the end of March 2025. (Source: OpenAI, via TechCrunch.) When discovery moves into AI answers at that scale, "are we cited?" becomes a number you want flowing into your own systems, not trapped in someone's dashboard.

The referral channel is climbing just as fast. Third-party measurements over autumn 2025 put ChatGPT referrals up roughly 52% year over year and Gemini referral traffic up around 388% in the same window (Similarweb data, via Digiday) — directional estimates, not our figures, but they all point the same way: AI is becoming a high-intent acquisition channel, and one most analytics stacks still misattribute. An API is how you wire that signal into the tools your team already runs.

Who it's for

The AI SEO API is for the people who want the data in their pipeline, not just on a screen:

• **Technical teams already in Claude Code** — wire SourceWatch in via MCP today and run the measure-and-act loop without a separate dashboard.
• **Agencies building client reporting** — pull per-client visibility and AI-referral data into white-label reports; the REST API (on the waitlist) is the clean fit for automated, scheduled exports.
• **In-house data & RevOps** — land AI-citation and share-of-voice metrics in your BI tool or warehouse alongside the rest of your acquisition data.
• **Builders & platforms** — anyone who wants AI-visibility data as a feed for their own product, once the REST surface opens.

If you're evaluating the broader product rather than just the integration, start at the AI visibility tracker hub or browse the best AI SEO tools roundup — the API is how you reach the same data programmatically. New to this? See how it works and the full feature set.

Get started

Two honest paths, depending on what you need:

1. 1

   Use MCP today

   Start a 14-day trial (card optional), connect a site, and add the SourceWatch MCP server in Claude Code or your MCP client. Your agent can read your visibility, citation and traffic data — and act on the gaps — within about 15 minutes of connecting.

2. 2

   Waitlist REST for later

   Need a language-agnostic HTTP endpoint for CI, a warehouse, or a dashboard? Join the REST API waitlist via /contact and tell us your stack. We'll bring you in as the read API opens; webhooks follow.

3. 3

   Check your starting point first

   Not sure you're even visible yet? Run the free single-page AI SEO audit — it checks whether AI engines can read and recognize one URL in about 15 seconds, no card required. A full-site scan runs on the trial.