Connect to the igeo.ai MCP server
igeo.ai offers two MCP endpoints: an anonymous public documentation server and an authenticated workspace server. Choose the public server for product research or setup questions, and the workspace server for authorized account data.
Public documentation server
Connect a Streamable HTTP MCP client to https://www.igeo.ai/api/docs-mcp/ without a key. The server card publishes the endpoint, publisher information, and supported protocol revision. A discovery mirror is available at /.well-known/mcp/server-card.json.
This stateless server implements the JSON response option of MCP 2025-11-25. It supports initialization, ping, tool discovery, and tool calls. It does not create sessions or expose an SSE subscription. After initialization, send MCP-Protocol-Version: 2025-11-25; POST requests use Content-Type: application/json and Accept: application/json, text/event-stream. Standard compatible clients handle these headers and the initialized notification.
{
"mcpServers": {
"igeo-public-docs": { "url": "https://www.igeo.ai/api/docs-mcp/" }
}
}
Five read-only tools expose public information: get_site_info, get_pricing, get_integration_docs, read_documentation, and search_documentation. Discover their current argument schemas using tools/list. For example, call search_documentation with {"query":"workspace authentication","limit":5}, then read the linked guide. Results identify their first-party sources. The server cannot access customer records, create keys, connect accounts, or publish content.
The official public agent integration repository contains a focused research skill, an Agent Plugins v1 manifest, and credential-free Claude Code and VS Code MCP configuration examples. Choose the format your client supports and merge the server entry into its existing configuration. The public research skill is also available directly from this site; its verified download is listed in the Agent Skills discovery index.
Requests are capped at 32 KiB and 240 requests per minute per running server instance. Rate-limit responses include Retry-After. Cross-origin browser requests require an explicitly allowed origin; ordinary server-side clients need no Origin header. No credentials belong on this public endpoint.
Authenticated workspace server
The workspace MCP endpoint accepts JSON-RPC 2.0 requests over HTTP at https://api.igeo.ai/mcp. Compatible clients can discover tools and resources, read workspace data, and call supported named write tools. MCP tool and resource access requires a workspace owner, admin, or editor role.
Configure your client
Create a key in MCP & API keys and copy the configuration generated for your workspace. For a client accepting an mcpServers configuration with HTTP headers, the shape is:
{
"mcpServers": {
"igeo": {
"url": "https://api.igeo.ai/mcp",
"headers": {
"Authorization": "Bearer <YOUR_API_KEY>",
"X-Workspace-Id": "<YOUR_WORKSPACE_ID>"
}
}
}
}
Client configuration formats vary. Use the client's remote HTTP MCP option and preserve both authentication values. Keep keys outside shared configuration files. Clients using hosted OAuth should follow the OAuth discovery details.
Initialize and discover
The following examples are JSON request bodies sent to POST /mcp with the authentication headers above and Content-Type: application/json.
{
"jsonrpc": "2.0",
"id": 1,
"method": "initialize",
"params": {
"protocolVersion": "2025-11-25",
"capabilities": {},
"clientInfo": { "name": "workspace-integration", "version": "1.0.0" }
}
}
Use the negotiated protocol version returned by the server. A standard MCP client handles its initialization lifecycle. Discover the current tool schemas next:
{ "jsonrpc": "2.0", "id": 2, "method": "tools/list", "params": {} }
Discover supported read routes through MCP itself:
{
"jsonrpc": "2.0",
"id": 3,
"method": "tools/call",
"params": { "name": "allowed_routes", "arguments": {} }
}
The same route catalog is available through resources/read using URI igeo://mcp/allowlisted-get-routes. Use resources/list to discover resources. The server also exposes a prompt-planning playbook at igeo://mcp/playbooks/prompt-prioritization.
Run a weekly planning read
{
"jsonrpc": "2.0",
"id": 4,
"method": "tools/call",
"params": {
"name": "prompt_work_priorities",
"arguments": { "recentDays": 30, "query": { "range": 7 } }
}
}
This read returns recently worked prompts, a maintenance pick, an improvement pick, and alternatives. recentDays controls the lookback for post activity; query.range controls the visibility window. The defaults are 30 days of activity and seven days of visibility. A pick can be empty. Describe only results returned for the selected workspace.
Useful follow-up tools include prompts, topics, competitors, prompt_responses, and posts. Use the discovered schema for arguments and retain returned IDs. Prefer named tools. The advanced api_get tool accepts a full path matching allowed_routes; it does not accept guessed bare paths.
Handle pagination and writes deliberately
For supported lists, omitting pagination lets MCP combine pages, with a cap of 50 upstream pages and 5,000 merged items. Check _mcpPagination for counts and truncation. Supply explicit pagination, such as query.take and query.skip, when you want one page.
Named writes include topic, prompt, tag, and competitor management, Brand Hub updates, and content generation or editing. Read existing state, review the proposed action, and use only tools discovered from the server. Check HTTP status, JSON-RPC errors, and tool isError before treating a call as successful. A JSON-RPC request ID correlates responses; it does not provide idempotency for mutations.
See authentication and retry guidance, the REST API guide, and product MCP documentation.