MCP Server Overview

iSearchFrom runs a hosted, first-party MCP server. Point any MCP-compatible AI client at it, authorize once, and your agent can run localized Google and Bing searches, manage rank-tracking and ad-tracking campaigns, read AI-visibility results, and query your organization data — all with permissions you control and enforced on every call.

The Model Context Protocol (MCP) is the open standard for connecting AI agents to external tools and data. Instead of copying results into a chat by hand, an MCP-connected agent discovers iSearchFrom’s tools automatically and calls them while it reasons — “check my brand’s visibility on ChatGPT for these ten prompts,” “run this search as a mobile user in Germany and summarise the ads,” “add a rank-tracking campaign for these keywords.”

Connection details

Everything an MCP client needs to connect:

Setting	Value
Server URL	https://api.isearchfrom.com/mcp
Transport	Streamable HTTP (stateless)
Auth	OAuth 2.1 — you authorize the client and pick which organization it acts in
Discovery	https://api.isearchfrom.com/.well-known/oauth-protected-resource

No API key, no local bridge

The MCP server uses OAuth, not an API key. You do not need to build a local bridge, run a proxy, or paste a secret into a config file — a compliant client discovers the auth endpoint from the server URL and walks you through a one-time browser sign-in. See Connect a client.

What your agent can do

The server exposes 34 tools across eight areas. Read tools return data; action tools create, change, or run things (and may spend tokens). Your agent only ever sees the tools your granted permissions allow.

Search

Run localized proxy and direct searches; read the parsed organic results, ads, and metadata.

Rank & ad tracking

Create, update, delete, and trigger position-tracker and ad-tracker campaigns; read run status.

AI visibility

Manage AI-visibility campaigns and read brand-mention metrics across ChatGPT, Gemini, Perplexity, and Google AI surfaces.

Projects

List, create, update, and delete the projects that group your searches and campaigns.

Flexible data queries

describe_schema + query_sql let the agent run read-only SQL over your own organization’s data, isolated at the database.

Usage & identity

whoami, token balance, entitlements, and token transaction history — so the agent can check budget before spending.

See the full Tool catalog for every tool, its purpose, and whether it costs tokens.

How authorization works

MCP access is user-scoped and least-privilege by design — this is a deliberate security property, not an afterthought:

1. You authorize a client once. The client opens a browser sign-in; you approve it and pick the organization it should act in.
2. You grant a subset of your own permissions. The client can never do more than you can, and you can grant it less. A read-only grant blocks every write tool automatically — the agent won’t even see them.
3. Every call is re-checked live. On each tool call the server intersects the consent-time grant with your current role permissions. If your access is reduced later, the client’s access drops with it immediately — no re-consent needed.
4. The organization is locked at the database. Read queries run as a non-owner database role with row-level security bound to your organization, so an agent can only ever see your own data.

Full detail — scopes, token spend, and how to revoke a connection — is in Permissions & security.

Prerequisites

• An iSearchFrom account and an organization with the MCP feature enabled on its plan. If your plan doesn’t include MCP access, the server returns a clear “feature not enabled” error at connect time. See Plans, tokens & limits.
• Enough token balance for the actions you intend to run. Searches and campaign runs consume tokens exactly as they do in the app — see Token usage.
• An MCP-compatible client. Connect a client has copy-paste setup for Claude Code, Claude Desktop, Cursor, Cline, and generic clients.

Next steps

Connect a clientCopy-paste setup for Claude Code, Claude Desktop, Cursor, Cline, and any MCP client.

Tool catalogEvery tool the server exposes, grouped by area, with token-cost notes.

Permissions & securityScopes, the consent flow, token spend, and revoking a connection.

REST APIPrefer to call the API directly from your own code? Start with the API overview.