Integrations

MCP Server

Use with Model Context Protocol-compatible AI clients.

Model Context Protocol gives assistants a structured way to request external context and invoke approved tools. The KALLISTRO MCP Server makes workspace, visibility, prompt, citation, and GEO audit data available to the clients your team already uses.

Quickstart

Connect your first client in a few minutes.

Tool reference

See the context and actions exposed to assistants.

Endpoint

Connect compatible clients to the streamable HTTP endpoint below.

HTTP

https://api.kallistro.ai/mcp

Frontend preview

This documentation page defines the intended connection experience. Enable the production endpoint and authentication flow when the MCP backend is provisioned.

Authentication

KALLISTRO uses an authorization flow that keeps credentials out of client configuration. When a client connects for the first time, it opens a browser sign-in and asks the user to approve a workspace scope.

No keys in config

Scoped access

Revocable sessions

Permission model

Read access and write actions are separated. A client only sees tools allowed by the connected account and selected workspace.

Quickstart

Add the remote server, complete sign-in in the browser, and confirm that KALLISTRO tools are available in your client.

Add the server

Use your client’s remote MCP configuration and enter https://api.kallistro.ai/mcp.

Authorize the workspace

Follow the browser prompt, sign in, and approve the workspace you want the assistant to access.

Ask a test question

Try: “List my available workspaces and summarize this week’s visibility changes.”

ChatGPT

In your connector settings, add a custom MCP connector and use the KALLISTRO endpoint. The browser authorization screen opens automatically when the connection is initiated.

Endpoint

https://api.kallistro.ai/mcp

Claude Code and Desktop

Claude Code can register the remote server from the terminal. Desktop clients can use the same endpoint through their connector interface.

Terminal

claude mcp add --transport http kallistro https://api.kallistro.ai/mcp

Cursor

Add a server entry to your MCP settings. Cursor opens the authorization flow the first time a KALLISTRO tool is requested.

JSON

{
  "mcpServers": {
    "kallistro": {
      "url": "https://api.kallistro.ai/mcp"
    }
  }
}

Codex

Register KALLISTRO as a remote MCP server in your Codex configuration, then complete browser authorization when prompted.

TOML

[mcp_servers.kallistro]
url = "https://api.kallistro.ai/mcp"

Available tools

Tools are organized around common revenue and AI visibility workflows. Names and response shapes below represent the intended public interface for the frontend documentation.

Tool	Required	Description
list_workspaces	—	List the KALLISTRO workspaces available to the connected account.
get_visibility_summary	workspaceId	Return coverage, trend, engine, and competitor visibility metrics.
list_monitored_prompts	workspaceId	List tracked buyer questions with intent, tags, and mention status.
get_prompt_response	promptId	Inspect model responses, brand mentions, ranks, and sentiment for one prompt.
list_citations	workspaceId	Find cited URLs and domains, filtered by engine, market, or date range.
get_geo_audit	auditId	Read crawlability, content readiness, structural issues, and recommended fixes.
list_recommendations	workspaceId	Return prioritized visibility actions with impact and confidence.

AI visibility

Visibility tools return coverage, ranking, sentiment, competitor, and AI-engine breakdowns for the approved workspace and date range.

Prompts and citations

Prompt tools connect buyer questions to model responses, brand mentions, and the pages or domains cited in those answers. Use them together to identify content gaps with traceable evidence.

GEO audits

Audit tools expose crawlability, AI-readiness, structural signals, dynamic-content checks, priority issues, and recommended fixes without requiring a report export.

Permissions

Read tools

Workspace discovery, reports, metrics, prompts, citations, audit results, and recommendations.

Write tools

Actions that create or change data. These only appear after explicit account and workspace approval.

Common arguments

Dates: use YYYY-MM-DD or a full ISO timestamp.
Markets: use lowercase ISO country codes such as us, in, or de.
Engines: filter by supported AI engines returned by the workspace metadata tool.
Pagination: list tools accept page and page-size parameters.

Troubleshooting

The client cannot connect

Confirm the endpoint, remote HTTP support, and that the production MCP service is enabled.

Authorization keeps reopening

Remove the existing connector session, reconnect, and approve the correct workspace.

A tool is missing

The account may not have the required permission or the selected workspace may not support that feature.

Ready to connect?

Create a workspace, then return to the quickstart.

Get started