Claude Plugin

What AIOProductOS/claude-plugin MCP server does

The AIOProductOS/claude-plugin MCP server exposes an AIOProductOS workspace through Model Context Protocol tools. It brings together customer feedback, account revenue, product analytics, support conversations, roadmap records, delivery work, releases, experiments, objectives, and codebase mapping. The central model is a product spine: customer evidence can be connected to features, tasks, outcomes, and strategic planning records.

The read side supports product discovery and analysis. Agents can inspect a product brain, generate or retrieve a weekly signal memo, compare planned work with shipped work, examine customer 360 records, and calculate NPS, revenue retention, funnel conversion, retention cohorts, and user paths. They can also browse ideas, initiatives, decisions, insights, features, objectives, experiments, releases, pages, sprints, tasks, conversations, channels, and artifact review versions.

The write side supports actions such as creating and updating tasks, promoting ideas, managing features and initiatives, planning sprints and releases, recording decisions, writing pages, capturing insights, posting to channels, and reviewing or reverting AI artifact versions. Actual access is constrained by the connected organization and the member’s permissions.

How it works

The AIOProductOS/claude-plugin MCP server can be used as a hosted connector at https://platform.aioproductos.com/api/mcp or through the documented npm stdio command. The hosted connection uses OAuth sign-in rather than an API key. The README describes OAuth 2.1 with dynamic client registration and PKCE, and says that sign-in occurs on first use when the connector requests workspace approval.

Several tools are intended to be called in a specific order. get_pm_playbook loads the product-management operating rules. whoami identifies the organization and available products, while pm_meta resolves names to IDs. Agents should use list or lookup tools before passing IDs to create, update, or comment operations. get_product_brain provides a starting snapshot, after which dedicated analytics and list tools can supply more detail.

Some operations have visible external effects. Replies and resolution actions affect the support inbox, channel posts appear to teammates, and booking changes notify meeting guests. delete_task is permanent, cascades to subtasks, and requires the exact confirmation value DELETE.

Setup and configuration

The AIOProductOS/claude-plugin MCP server requires an AIOProductOS workspace. The project documents installation as an official Claude Code plugin and also provides an npm-based stdio connection for other MCP hosts. The README states that the plugin itself is free but drives a paid AIOProductOS workspace; it also mentions a 14-day onboarding runway for new organizations.

For the hosted connector, connect the MCP client to the documented endpoint and complete the OAuth sign-in flow. No API key is listed as required. For local stdio use, run the project’s published package with the package runner shown in the README. Workspace data remains scoped to the organization and the connected role.

Tools and capabilities

Key capability groups include:

• Ground product reasoning in customer signals, revenue, usage, objectives, and the product brain.
• Analyze NPS, NRR, conversion funnels, retention cohorts, and post-event paths.
• Review customers, support threads, insights, ideas, features, pages, experiments, and releases.
• Manage tasks, comments, sprints, initiatives, objectives, decisions, and roadmap records.
• Capture insights, vote on ideas, promote ideas into features, and connect records across the product spine.
• Inspect codebase maps and versioned AI artifact reviews.

The repository also documents interactive MCP Apps for task boards, weekly signal views, and funnel results when the host supports MCP Apps. Hosts without that support receive the regular text tool responses instead.