iSearchFrom

npx add-mcp 'https://api.isearchfrom.com/mcp'

Your AI assistant, now fluent in Google.

Connect Claude, ChatGPT, Cursor (or any AI assistant that supports MCP) and it can run localized searches, track your rankings and pull reports for you. No more copy-pasting between tabs: paste one link, approve access in your browser, done.

Works with Claude, ChatGPT, Cursor & more You approve exactly what it can access Nothing to install, connected in a minute

New to MCP? Start with the basics

SELECT keyword, position
FROM serp_position_results
ORDER BY position;

4 rows · 38ms · read-only

get_token_balance

1,240

resettable · resets in 12d

Claude Connected to iSearchFrom

Where does runfaster.com rank for "running shoes" in Berlin on mobile?

run_proxy_search isearchfrom · mcp

Completed

{ "q": "running shoes", "country": "DE", "city": "Berlin", "device": "mobile" }

1 page scraped · 1 token via /mcp

1

Running Shoes & Sneakers nike.com

2

Laufschuhe online kaufen zalando.de

3

Running Shoes · Top Angebote amazon.de

4

Run Faster · Performance Shoes runfaster.com

You're #4 on mobile in Berlin, up two spots since last week. Nike holds #1, with two marketplaces right behind.

Reply to Claude… isearchfrom

In the wild

Four jobs you can delegate this afternoon.

A few of the workflows an MCP-connected agent handles end to end.

#seo 7:30 AM

Claude app

3 keywords up, 1 slipped · full brief in #seo

running shoes	#4	2
trail runners	#7	3
marathon gear	#11	1
racing flats	#9	2

3 2

The morning rank report

query_sql

A scheduled agent pulls the week's position checks, writes the summary and posts it before you're at your desk.

iSearchFrom now

New advertiser on "crm software"

Sponsored

P pipeflow.io

CRM Software · Free 30-Day Trial

Automate your pipeline in minutes. No credit card required.

Competitor ads, on watch

trigger_ad_tracker_run

The agent runs the ad tracker, compares the results and pings you the moment a new competitor bids on your keywords.

landing-page.tsx · Cursor

› where do we rank for "running shoes" in Berlin, London and New York?

run_proxy_search × 3

#4 #7 #12 ▍

SERP checks from inside the IDE

run_proxy_search

Ask Cursor or Claude Code how a page ranks in three markets mid-review. The answer lands in the same chat.

Brand mentions 2/3 mentioned

Mentioned

48%

Mentioned

34%

Missing

5%

AI visibility audits on demand

trigger_ai_visibility_run

One prompt sweeps ChatGPT, Gemini and Perplexity for your brand: mentions, sentiment and who's recommended instead.

Works with every MCP-enabled client

One endpoint. Every agent.

Six clients pictured, hundreds supported. Anything that speaks MCP plugs into the same URL.

Claude Claude Code ChatGPT

MCP endpoint /mcp api.isearchfrom.com/mcp

Cursor

Windsurf

VS Code

…and any other MCP client

89

agent tools

1 click

to revoke access

1 URL

to connect

0

installs required

The toolbox

Everything the app does, your agent can do too.

9 tool families cover the whole workflow (searches, rank, ad, Maps, marketplace and AI-answer tracking, projects and usage), plus a way to ask questions across all your data.

8 tools

Localized search

Run Google or Bing searches from any country, city, language and device: full scraped result pages or direct search URLs.

run_proxy_search run_direct_search get_search_options

Ask your data anything

Ask in plain language ("which keywords dropped this month?") and your agent turns it into a safe, read-only query across your rankings, ads, mentions and searches. It can read everything in your workspace and change nothing.

SELECT keyword, MIN(position) AS best
FROM serp_position_results
WHERE checked_at > now() - '7 days'
GROUP BY keyword
ORDER BY best ASC;

read-only · rls scoped 4 rows · 38ms

query_sql describe_schema

12 tools

Position tracking

Create campaigns, trigger runs and read keyword rankings by location.

create_position_campaign get_position_results

12 tools

Ad tracking

Watch competitors' paid placements: campaigns, runs and per-keyword ad captures.

create_ad_tracker_campaign get_ad_tracker_results

17 tools

AI visibility

Track brand mentions across ChatGPT, Gemini and Perplexity answers, on schedule or on demand.

create_ai_visibility_campaign get_ai_visibility_results

14 tools

Maps visibility

Geo-grid rank on Google Maps: run a grid, read the heatmap and see which competitors hold each point.

run_maps_campaign get_maps_heatmap

15 tools

Marketplace visibility

Shelf position on Amazon and Google Shopping: rank, price and sponsored share for your products and the listings around them.

get_marketplace_overview get_marketplace_competitors

5 tools

Projects

List, create and organize projects so agents keep work tidy across clients.

list_projects create_project

4 tools

Usage & budget

Balance, plan limits and the full usage history, so agents can budget their own token spend.

whoami get_token_balance get_token_transactions

60-second setup

From zero to agent-ready in three steps.

Nothing to install, no code to write, no API keys to juggle. Point your AI app at one link, approve access in your browser, and the tools show up on their own.

1 Step 1

Paste one link into your AI app

Claude, ChatGPT, Cursor and every other MCP client accept a remote server URL. Paste ours once. There's nothing to install and nothing to keep running.

In the app:

1. 1 Open Settings → Connectors
2. 2 Choose "Add custom connector"
3. 3 Paste the server URL and connect

https://api.isearchfrom.com/mcp

connected · 89 tools available

2 Step 2

Approve exactly what it can touch

Your browser opens a consent screen. Pick the workspace and tick only what you're delegating: read searches, run searches, manage campaigns. What you approve is all the agent ever gets, and you can revoke it anytime.

app.isearchfrom.com/oauth/consent

Claude wants to connect

to your iSearchFrom workspace

Workspace

Acme Marketing

Read searches & results

Run new searches

Read projects

Manage & delete campaigns

Deny Approve access

You can revoke this connection at any time

Guardrails

Scoped by you. Isolated by design.

Every tool call clears four gates before it touches a row of data. Agents get exactly the reach you granted, nothing more, and never another workspace's data.

Agent request

01

You approve access first

Permissions are granted per connection on an OAuth 2.1 consent screen, and frozen at approval.

02

Live permission check

Every call re-checks your current team role. Downgrades apply instantly.

03

Row-level security

Queries are pinned to your workspace at the database level.

04

Read-only SQL sandbox

A non-owner role, SELECT-only grants, row caps and strict timeouts.

Your data

Revoke a connection with one click

Write access is opt-in, never default

No system tables, no other workspaces, no writes

Full audit trail in the usage ledger

Questions

Frequently asked questions

Model Context Protocol is the open standard AI assistants use to call external tools. Connecting iSearchFrom over MCP means Claude, Cursor, ChatGPT and other agents can run searches, manage campaigns and query results for you, instead of you copy-pasting between tabs.

Any MCP client that supports remote servers: Claude (web, desktop and Claude Code), Cursor, Windsurf, ChatGPT (with developer mode enabled), VS Code and more. If your tool speaks MCP, it can connect.

The MCP server is included in the Team plan. It's separate from API access, and unlike the API there's nothing to code: you connect straight from your AI app's settings.

There's no separate metering. Tool calls spend the same token balance as the app: a proxy search costs one token per page, tracker runs one token per keyword or check. All reads (results, analytics, SQL) are free.

Only if you let it. Write and delete capabilities are individual checkboxes on the consent screen. Grant read-only access and the agent can look but never touch.

Every connection is bound to one workspace. Row-level security pins queries to that workspace at the database layer, the SQL tool runs on a read-only role with column-level grants, and system tables are unreachable.

Revoke the connection at any time. The token dies immediately. Role downgrades propagate too, since permissions are re-checked live on every call.

MCP server

Put an agent on your search ops.

Connect your assistant in under a minute and hand off the busywork (localized checks, tracking runs, reports) while you keep full control of what it can touch.

Works with Claude, Cursor, ChatGPT and any MCP client

One link to connect. You approve access in your browser

Ask questions across your rankings, ads, map coverage, marketplace shelf and mentions

Included in the Team plan

Your MCP endpoint

https://api.isearchfrom.com/mcp

89

agent tools

1 click

to revoke access

< 1 min

to connect

100%

your workspace only