Connecting to the Poggio MCP Server

Overview

This guide walks enterprise IT administrators through exposing Poggio's capabilities as an MCP server, enabling external AI clients to access your organization's knowledge base and agent capabilities. By connecting external MCP clients to Poggio, you can extend your AI infrastructure beyond the Poggio interface while maintaining centralized knowledge management.

Business impact

  • Unified AI ecosystem: Provide consistent access to organizational knowledge across multiple AI interfaces

  • Developer productivity: Enable engineers to access Poggio's capabilities directly via MCP in order to accelerate and enhance agent development

  • Flexible integration: Support custom applications and workflows that access to deeply researched account intelligence and tailored POVs

  • Centralized knowledge: Maintain a single source of truth for account intelligence while enabling distributed access

Prerequisites

Setup

  • Point your client to https://mcp.poggio.io/mcp.

  • Upon connecting, you will redirected to the Poggio app to authorize the new client.

authorize
Example for Connecting to Claude

Available Poggio capabilities via MCP

When connected to Poggio’s MCP server, external clients can use these tools:

  • search: Returns a list of account intelligence documents for a given query. The query must begin with a valid domain, e.g. "example.com" or "example.com product strategy".

  • fetch: Returns the full text and metadata for a specific document ID. Use IDs returned by search. IDs include workspace-level workspace_info and page IDs like account:<company_id>:page:<page_id>.

  • create_account: Creates a new account in your Poggio workspace using a domain, e.g. create_account(account_domain="example.com"). Intelligence generation begins immediately and becomes available via search / fetch within minutes.

  • recreate_account_pages: Recreates all pages for an existing account using current default templates. This deletes all existing pages and creates fresh pages from current templates.

Note: Search queries without a leading valid domain will result in an error.

Tool Reference

search

  • Parameters: query (string). Must start with a valid domain (e.g., "example.com"). Optional free text can follow the domain.

  • Returns: List of Item objects with fields:

    • id (string)

    • title (string)

    • text (string snippet)

    • url (string)

  • Example result:

  • Errors: bad request if the query lacks a valid leading domain; not found if no account exists for the domain.

fetch

  • Parameters: id (string). Use an id returned by search. Accepted formats include workspace_info and account:<company_id>:page:<page_id>.

  • Returns: ItemWithMetadata with fields:

    • id

    • title

    • text (full markdown)

    • url

    • metadata

  • Example result:

  • Errors: bad request for malformed IDs; not found if the page or workspace item is unavailable.

create_account

  • Parameters: account_domain (string). Example: "example.com".

  • Returns: Object containing:

    • id (account id)

    • org_id

    • domain

  • Example result:

Note: Triggers deep research for the domain; resulting resources are generally available via search / fetch within minutes.

recreate_account_pages

  • Parameters: account_domain (string). Example: "example.com".

  • Returns: Object containing:

    • id (account id)

    • message

    • domain

  • Example result:

Note: Triggers deep research for the domain; resulting resources are generally available via search / fetch within minutes.

Security Best Practices

Token management

  • Generate separate tokens for each client or use case

  • Use descriptive names to track token purposes

  • Implement token rotation policies aligned with your security requirements

Access control

  • Limit MCP server access to authorized clients only

  • Monitor token usage patterns for unusual activity

  • Implement network-level restrictions where possible

  • Regular security audits of client configurations

Data protection

  • Understand that MCP clients will have access to workspace data

  • Ensure client applications meet your organization's security standards

  • Implement appropriate data handling policies for external clients

  • Monitor data access patterns and usage

Troubleshooting common issues

Connection failures

  • Symptom: Client cannot connect to Poggio MCP server

  • Solution:

    • Verify the server URL is correct and accessible

    • Check that the API token is valid and not expired

    • Confirm network connectivity and firewall settings

    • Ensure the MCP Server feature is enabled for your workspace

Authentication Errors

  • Symptom: "Invalid token" or authentication failures

  • Solution:

    • Verify the API token is copied correctly without extra spaces

    • Check that the token has not expired

    • Confirm the token has not been revoked

    • Generate a new token if necessary

Performance Issues

  • Symptom: Higher than expected latency or API quota back-pressure from Poggio MCP server

  • Solution:

    • Check network connectivity and latency

    • Monitor response codes, paying special attention to 429: Too Many Requests responses.

    • Consider implementing caching strategies in client applications, and/or reach out to Poggio support to request higher API request quotas.

Next Steps

After successfully connecting MCP clients to Poggio:

  1. Scale integration: Connect additional client applications and development tools

  2. Custom development: Build custom applications leveraging Poggio's MCP capabilities

  3. Team training: Educate developers on effective use of Poggio via MCP clients

For additional assistance, contact Poggio support or your account manager.

Last updated