Google Discover MCP Server

Read the Google Discover feed of a country and check if your domain is in it.

Hosted MCP Server

npx add-mcp 'https://cotlet.dev/mcp/v1/google-discover'

Installs into Claude Code, Codex, Cursor and more

Documentation

Cotlet · Integrations · How it works · Security · For developers · For automation agencies · Blog · API docs · Help center · Changelog · Contact · Pricing · Sign up

Google Discover API and MCP server: See what Google Discover shows in any country.

Google has no official MCP server or API for the Discover feed (checked 25 Oct 2026). Cotlet runs 3 hosted Google Discover tools that read the feed of a country, over MCP for agents and REST for scripts.

Read Google Discover country feeds and check domain appearances through MCP or REST. Use an AI agent, n8n, Make, Zapier or a script. No Google account or Cloud Browser is needed.

Connect an AI agent Use the REST API

Free 7-day trial, no card
No Google account needed
Unofficial, not affiliated with Google
Page: https://cotlet.dev/integrations/google-discover-api-mcp-server

What can you automate in Google Discover?

Read the feed of a country
Check if your story is in Discover
Watch a domain every hour
Compare countries and languages
Count competitors' cards
List the videos Discover pushes

Does Google Discover have an MCP server or an API?

Google's only official Discover data is the Discover report in Search Console: clicks and impressions for sites you own, not which cards a country sees.
The feed lives in apps: Chrome's New Tab page on phones, the Google app and Pixel phones. Mobile google.com showed only trending searches in our checks on 25 Oct 2026.
Paid panel tools such as DiscoverSnoop sell totals from the personal feeds of many users. We found no GitHub MCP server for the feed (searched 25 Oct 2026).

What you need	Google's own data	Panel tool	Cotlet
The cards a country sees now, in order	No	Totals	Yes, each card
Is my domain in the feed right now?	No	In its panel	Yes, per call
Any country, any time, from code	No	Paid API	Yes, MCP and REST
No Google account needed	N/A	Yes	Yes
Clicks and impressions of your site	Yes	Panel counts	No

Checked 25 Oct 2026. Sources: Search Central: Google Discover, Search Console Help: Discover performance report, Chromium source: the New Tab feed (components/feed), DiscoverSnoop pricing

How does Cotlet connect to Google Discover?

Activate. In the Cotlet console, one click.
Choose MCP or REST. Connect your agent or create an API key.
Run your workflow. Ask your agent or call the REST API.

No Google account is used, so nothing of yours is signed in or stored.

Connect Google Discover to your AI agent or workflow

Use MCP with ChatGPT, Claude, Cursor or Codex, or authenticated HTTP in n8n, Make or Zapier. The REST guide follows the agent setup.

Activate Google Discover. Sign up, open Google Discover under Integrations in the console and press Activate. There is no Google account to connect.
Paste the Cotlet address. Add https://cotlet.dev/mcp/v1/google-discover as an MCP server or connector in your agent.
Sign in or paste a key. Chat apps sign you in to Cotlet; editors, n8n and scripts take an API key from the console.

https://cotlet.dev/mcp/v1 serves every app you have access to, Google Discover included. Got access after connecting? Refresh the connector.

Claude:

Settings > Connectors > Add custom connector
URL: https://cotlet.dev/mcp/v1/connect

Claude Code:
claude mcp add --transport http cotlet-google-discover \
  https://cotlet.dev/mcp/v1/google-discover \
  --header "Authorization: Bearer $COTLET_API_KEY"

ChatGPT:

Settings > Apps & Connectors > Create
MCP server URL: https://cotlet.dev/mcp/v1/connect
Authentication: OAuth

Cursor:

// ~/.cursor/mcp.json
{
  "mcpServers": {
    "cotlet-google-discover": {
      "url": "https://cotlet.dev/mcp/v1/google-discover",
      "headers": { "Authorization": "Bearer ${env:COTLET_API_KEY}" }
    }
  }
}

Codex:

# ~/.codex/config.toml
[mcp_servers.cotlet-google-discover]
url = "https://cotlet.dev/mcp/v1/google-discover"
bearer_token_env_var = "COTLET_API_KEY"

n8n:

MCP Client Tool (inside an AI Agent)
Transport: HTTP Streamable
URL: https://cotlet.dev/mcp/v1
Auth: Bearer, <your key>
Tool: google_discover_read

Make, Zapier:

Make: MCP Client > Call a tool
Zapier: MCP Client by Zapier > Run Tool
Server URL: https://cotlet.dev/mcp/v1
Token: <your key>
Tool: google_discover_read

How to use the Google Discover REST API

Create your Cotlet API key

Sign up to Cotlet, request or enable Google Discover access in the console, and create a Cotlet API key scoped to Google Discover reads. No Google account, cookies, sign-in or Cloud Browser is needed.

Save your key as COTLET_API_KEY in your environment. Send Authorization: Bearer <your Cotlet API key>; never use an app password or cookie as the key.

Make your first Google Discover API request

This GET request reads data without changing it.

cURL: first read:

curl "https://cotlet.dev/api/v1/google-discover/feed?country=US&limit=5" \
  -H "Authorization: Bearer $COTLET_API_KEY"

A successful REST response puts the tool output in result. This example response excerpt omits call IDs, messages and credit details.

Example response excerpt:

{
  "status": "success",
  "app": "google-discover",
  "tool": "get_discover_feed",
  "result": {
    "country": "US",
    "language": "en-US",
    "fetched_at": "2026-10-30T12:00:00+00:00",
    "cards": [
      {
        "position": 1,
        "title": "Example story",
        "publisher": "Example News",
        "url": "https://example.com/story",
        "domain": "example.com",
        "type": "article",
        "age": "3h"
      }
    ]
  }
}

Read the sampled feed

There is no cursor. limit trims the top cards of one feed, usually around 30 cards, with a maximum limit of 50.

A country and language feed is cached for up to 15 minutes.

This is the signed-out Chrome for Android default country feed, not a personal feed, traffic or Search Console impressions. Five cards do not represent complete Discover coverage.

check_domain_in_feed reads GET /feed/domains/{domain}, with optional country and language, including subdomains. in_feed false means no matching card in the current sampled default feed; it does not prove absence from all Discover.

Google Discover workflows with n8n, Make and Zapier

Tool	Connect through	Example workflow
n8n	HTTP Request	Check a domain on a schedule, save each result with a timestamp and alert on new cards.
Make	HTTP > Make a request	Compare country and language feed snapshots in an editorial report.
Zapier	Webhooks by Zapier > GET or Custom Request	Save new card URLs to a research queue, skipping URLs already handled.

Configure the schedule, destination and saved state in your automation tool. Cotlet reads the source; your configured destination step writes the report or sends the alert.

Compare existing record state as well as IDs when detecting changes. Python and Node.js can use the same HTTP endpoints without an SDK or agent.

Endpoint and tool reference. The REST API guide covers requests, errors and approvals.

Google Discover API endpoints and MCP tools

The feed of a country

About 30 cards per read, in feed order. Country and language per call; one read is reused for 15 minutes.

Read the Discover feed (get_discover_feed, read): The cards Google Discover shows right now in a country, in feed order, with no Google account.
Check if a domain is in the Discover feed (check_domain_in_feed, read): Whether cards from a domain are in a country's Discover feed now, and at which positions.

Many cards at once

Each read draws its cards from a larger pool. One call reads the feed many times from different home IPs and returns up to 300 different cards, with how often each appeared. A call costs 100 credits, because it makes many reads.

Collect Discover cards (collect_discover_cards, read): Up to 300 different cards Discover shows in a country, from many reads, with how often each appeared.

Inputs, REST path and example response of every tool: https://cotlet.dev/integrations/google-discover-api-mcp-server#tools

Which Discover feed do the tools read?

Whose feed. The one Chrome for Android shows on its New Tab page to someone not signed in to Google.
Where from. A home IP in the country you pass, a fresh one for each read.
What each card has. Position, title, publisher, URL, domain, article or video, age and image.
How fresh. Read on the call, reused for up to 15 minutes per country and language.
What it is not. Not your personal feed, and not totals from a panel of users.
What we keep. Time, tool and status of each call. No Google login, no cookie.

Local news follows the city of the exit: a UK read on 25 Oct 2026 held Liverpool and Manchester outlets, a US read a West Virginia TV station. Both reads held 31 cards; 12 of the US cards and 6 of the UK cards were YouTube videos.

Is it against Google's terms?

using automated means to access content from any of our services in violation of the machine-readable instructions on our web pages (for example, robots.txt files that disallow crawling, training, or other activities)

Google Terms of Service, effective 30 Jul 2026

Google lists this among the ways to abuse its services. The tools call the same feed endpoint Chrome calls for a signed-out user, with Chrome's own key. Read Google's terms before you use them: you are responsible for how you use the tools and the data they return. If Google blocks the calls, the tools return an error and the calls cost nothing.

What you should know:

No Google account of yours is used, so none can be suspended
Read only: the tools change nothing at Google
Each read leaves from a home IP in the country you ask for
Repeat reads within 15 minutes come from our cache, not from Google

What happens when Google changes the feed?

12 h between tests of the US and UK feeds and a domain check
15 min that one read of a country and language is reused
4/4 checks passed on Nov 6, 2026
$0 for a call that fails

An empty feed or a refused request fails the test and the call, instead of returning an empty list as if your domain were absent. status.cotlet.dev checks the site, console and API every 3 minutes.

One domain check, end to end

Reads the feed. The UK feed, from a home IP in the UK, about 30 cards.
Keeps your domain. Cards on example-news.com or its subdomains.
Reports the place. Position, title and age of each, and the feed size.
Runs again. Every hour, so you see when the story enters and leaves.

MCP call:

{
  "name": "check_domain_in_feed",
  "arguments": {
    "domain": "example-news.com",
    "country": "GB"
  }
}

REST:

curl "https://cotlet.dev/api/v1/google-discover/feed/domains/example-news.com?country=GB" \
  -H "Authorization: Bearer $COTLET_API_KEY"

Tool output:

{
  "domain": "example-news.com",
  "country": "GB",
  "language": "en-GB",
  "fetched_at": "2026-10-25T16:29:28+00:00",
  "in_feed": true,
  "feed_size": 31,
  "cards": [
    {
      "position": 4,
      "title": "Ten herbs that survive a cold windowsill",
      "publisher": "Example News",
      "url": "https://example-news.com/guides/windowsill-herbs",
      "domain": "example-news.com",
      "type": "article",
      "age": "2h",
      "image_url": "https://encrypted-tbn0.gstatic.com/images?q=tbn:example"
    }
  ]
}

in_feed is false when no card of the domain is in the feed at that moment; feed_size says how many cards the feed held. REST under /api/v1/google-discover has the same tools, one key, one price. The tools table lists the REST path of each tool.

How much do the Google Discover API and MCP server cost?

About $52 a month for one domain checked every hour, 9,000 credits.

Job	Calls	Credits
Check your domain in the US feed, hourly	1 × 720	7,200
Read the US, UK and German feeds, daily	3 × 30	900
Check your domain in 3 more countries, daily	3 × 30	900
Total	900	9,000

Free 7-day trial, no card
10 credits per successful call, MCP or REST
No Cloud Browser day: nothing to keep signed in
One-time packs, valid 12 months, no subscription

The volume is our assumption for one publisher watching its own site. Dollars at the $29 pack rate.

Google Discover data by hand, from a panel or hosted: which to use

Option	What it does	Where it stops
By hand on a phone	Open Chrome's New Tab page or the Google app and scroll.	One country, the one your phone or VPN is in; signed in it shows your own feed; nothing is recorded.
DiscoverSnoop	Rankings, trends and publisher stats from a panel of users' feeds in 54 countries.	Totals, not the feed one person sees; 5 countries on the $299 a month Pro plan, API $299 a month on top or in the $2,999 Full Access plan.
NewzDash	Discover tracking inside a news SEO dashboard.	Plans after a demo; built for newsrooms, not for an agent call.
GitHub	We found no MCP server or open script for the feed (searched 25 Oct 2026).	Nothing to run.

Pick a panel tool if you need history and market share across many users. Pick Cotlet if your agent should look at the feed of a country right now, as often as you like, and tell you where your story is.

Run our Google Discover MCP server on your own computer

All 3 Google Discover tools are also public on GitHub, as an MCP server and REST API in Python. It sends the same request as Chrome for Android, from your own connection, so you can see the feed of your country and how each tool reads it. You take Chrome's API key from a Chrome APK yourself, and the code is a trimmed version of what runs on cotlet.dev, free to use in your own business.

Free for personal use, learning and testing
Reads the feed of the country your IP is in; other countries need a proxy of your own
You renew the key and update the code yourself when Google changes the feed
License Elastic 2.0: free in your own business; ChatGPT, your phone and 24/7 runs need hosted Cotlet
Code: https://github.com/cotletdev/google-discover-api-mcp-server

Also listed on: Odel, mcpbeat

Who runs Cotlet

Cotlet is run by Take Group sp. z o.o., a software company registered in Poland, EU (VAT PL6342984273), working since 2019. The founders are Bart Krawczyk (Co-founder, CTO) and Jake Slugocki (Co-founder, SEO, sales and marketing). Questions go to contact@cotlet.dev.

Google Discover API and MCP server: questions

Can I use the API without an AI agent?

Yes. Use a Cotlet API key scoped to Google Discover and authenticated HTTP. No Google account or Cloud Browser is needed. Start with the REST guide.

Can I use n8n, Make or Zapier?

Yes. Use authenticated HTTP requests with your Cotlet key, or an MCP client for agent workflows. You configure the schedule, destination and record state. See app-specific workflows.

How do I get access?

Sign up free at https://cotlet.dev, open Google Discover under Integrations in the console and press Activate. There is no Google account to connect. Your free trial starts when you sign up, no card needed.

Do I need a Google account?

No. The tools read the feed someone sees who is not signed in to Google, so there is nothing to sign in to, and no Cloud Browser runs for this app.

How do I get Google Discover into ChatGPT or Claude?

Sign up at cotlet.dev and add Cotlet to ChatGPT or Claude, following the steps in your console. No API key and no Google account are needed. ChatGPT and Claude then read the live Discover feed through Cotlet, in their phone apps too.

What can ChatGPT do in Google Discover?

With Cotlet, ChatGPT, Claude and other AI agents handle jobs like these in Google Discover: see what Google Discover shows in a country right now; check if your story entered Google Discover; monitor your domain in Discover across countries. You ask in plain words, and anything that reaches other people waits for your OK first.

Is this the same data as DiscoverSnoop or another Discover panel?

No. A panel collects the personal feeds of many users who run its software and sells totals: rankings, clicks and history. Cotlet reads one feed per call, the default feed Chrome for Android shows to someone not signed in, from a home IP in the country you pass. It has no click counts and no history; your agent builds the history by calling on a schedule.

Is it my personal Discover feed?

No. No Google account is involved, so your searches do not shape it. People who are signed in see feeds tuned to their activity, so their cards differ from these.

Why does the feed of one country change between calls?

Discover changes its feed through the day, and each read leaves from a different home IP in the country, so local news follows that city: a UK read on 25 Oct 2026 held Liverpool and Manchester outlets. Reads of one country and language are reused for up to 15 minutes, so calls inside that window return the same cards and the same fetched_at.

Which countries and languages work?

Any two-letter ISO country code where we have a home IP. The country decides the region, language (BCP 47, such as de-DE) the language of the stories, by default the country's main one. Reads from the US, GB, DE, FR, PL, BR, JP and IN all answered on 25 Oct 2026.

What can it not do yet?

It reads the signed-out feed of a country and checks a domain in it, nothing else. It does not open the more-stories group of a card or read the feed of a signed-in person, and it keeps no history: one call is one look at the feed.

Is the MCP server different from the REST API?

Only the way in. Both have the same 3 tools, inputs and answers, one API key and one price. Agents use MCP; scripts, cron jobs and HTTP steps use REST.

How many Discover cards can I get?

One read holds about 30. Each read draws them from a larger pool, so collect_discover_cards reads the feed many times from different home IPs and merges them: up to 300 per call, for 100 credits. In our US test on 25 Oct 2026, 300 different cards took 88 reads and 55 seconds; complete says if the feed held fewer than you asked for.

Is Cotlet affiliated with Google?

No. Cotlet is independent and not affiliated with Google.

Give your AI agents your signed-in browser.

Product: Integrations · How it works · Pricing · Security

Resources: Blog · API docs · Help center · Changelog · GitHub

Solutions: For developers · For automation agencies

Company: About · AI info: company facts · Contact

Legal: Terms of service · Privacy policy · Refund policy · Acceptable use · Cookie policy · DPA · Subprocessors

Elsewhere: Status page · Cotlet on GitHub · Cotlet on X · Cotlet on LinkedIn · Cotlet on Product Hunt · Cotlet on Crunchbase · Cotlet on SaaSHub

Cotlet.dev is built by Bart Krawczyk and Jake Slugocki, the founders of Take Group sp. z o.o., ul. Wincentego Janasa 4, 40-855 Katowice, Poland. KRS 0000871398, NIP 6342984273, VAT EU: PL6342984273. About us

Cotlet is an independent service. It is not affiliated with, endorsed by or sponsored by any of the products listed on this site. Product names are trademarks of their respective owners.