Skip to main content
Glama
README.md
# AIOProductOS plugin for Claude Code

[![claude-plugin MCP server](https://glama.ai/mcp/servers/AIOProductOS/claude-plugin/badges/card.svg)](https://glama.ai/mcp/servers/AIOProductOS/claude-plugin)

[![Hosted connector on Glama](https://img.shields.io/badge/Glama-verified_connector-2ea87c)](https://glama.ai/mcp/connectors/com.aioproductos/mcp) [![MCP Registry](https://img.shields.io/badge/MCP_Registry-com.aioproductos%2Fmcp-4b5563)](https://registry.modelcontextprotocol.io/v0/servers?search=com.aioproductos)

Connect your [AIOProductOS](https://aioproductos.com) workspace to Claude Code and work your whole product from chat — feedback, revenue, tasks, releases, analytics, and code activity joined on one customer record.

One install gives you:

- **The workspace connection** — the AIOProductOS remote MCP server: **71 tools over the product spine**, and it doesn't just read. It's a fully autonomous PM surface — read the spine _and_ write to it across the whole strategy-to-delivery ladder: capture and vote on ideas, promote them into features, set initiatives and objectives & key results, plan sprints and releases, run experiments, log decisions, write pages, and create/update/delete tasks — plus post to team channels, capture insights, and merge/unmerge people. OAuth sign-in, no API keys.
- **`/aioproductos:triage`** — Monday-morning product triage: surfaces the top 3 customer-evidenced insights, scores them transparently from real revenue and usage, and (after you confirm) creates the tasks, assigns a named AI teammate, and schedules them into the next sprint.
- **The `aioproductos-triage` skill** — the full triage doctrine, available whenever you ask "what should we work on next?"

## What you can do

**Read the spine** — customer 360, product brain, weekly signal memo, roadmap drift; analytics: NPS, NRR, funnels, retention, path analysis; list ideas, initiatives, tasks, features, objectives, sprints, releases, pages, experiments, insights, conversations, channels, bookings, identity merges, codebase map.

**Write to the spine** — no jumping back to the web UI:

- **Discovery** — capture ideas, vote on them, promote an idea into a feature
- **Strategy** — create/update initiatives, objectives and key results, decisions, experiments
- **Roadmap** — create/update features, releases, pages
- **Delivery** — create/update tasks, comment, create/update sprints
- **Customers & comms** — capture insights, post to channels, merge/unmerge end users
- **AI artifacts** — review artifacts, revert to a prior version

Everything is scoped to your organization and your role's permissions, and writes go through the same guardrails as the app.

**Writes someone else can see.** Three groups reach beyond your own workspace, so
they're worth knowing about before you connect: `reply_to_conversation`,
`add_note` and `resolve_conversation` act on your support inbox and a visitor
sees the reply; `post_to_channel` and `reply_in_channel` post to team channels as
you; `cancel_booking` and `reschedule_booking` change a booked meeting and notify
the guest.

**One destructive tool.** `delete_task` is permanent, cascades to subtasks, and
has no undo — so it requires `confirm: "DELETE"` alongside the task id. A task id
alone can never trigger it; the assistant has to ask you first. Nothing else
deletes anything.

## Interactive apps

Three tools render a working interface in the conversation rather than a wall of JSON
([MCP Apps](https://modelcontextprotocol.io/extensions/apps/overview), SEP-1865). They aren't
screenshots — each one calls back into this same MCP:

| Ask for | You get | What it writes |
|---|---|---|
| `list_tasks` | **Board** — your tasks grouped by status, ranked by priority | change a status → `update_task` |
| `get_weekly_signal_memo` | **Weekly signal** — this week's themes with the evidence behind each | a theme → `create_task`, linked to its feature |
| `analyze_funnel` | **Funnel** — conversion per step with the live MRR behind each one | read-only |

Hosts that support MCP Apps (Claude, ChatGPT, VS Code, Goose and others) render them inline.
Everywhere else the tools return exactly the text they always did — the app is an enhancement,
never the only copy of the answer.

## Install

```
/plugin marketplace add AIOProductOS/claude-plugin
/plugin install aioproductos@aioproductos
```

On first use, the AIOProductOS connector will prompt you to sign in with your workspace account and approve access. All data is scoped to your organization and your role's permissions.

## Requirements

An AIOProductOS workspace ([aioproductos.com](https://aioproductos.com)). New organizations get a 14-day onboarding runway on their own data, and every plan carries a 30-day money-back guarantee. Your AI usage is never metered — flat plans, no credit meters.

## What the triage looks like

```
This week's triage — top 3 (evidence-backed). Confirm before I create them.

# | Insight (source)              | Proposed task            | Score
--+-------------------------------+--------------------------+------
1 | "Need SAML" — 3 accts, $4.2k  | SSO/SAML for enterprise  | 2.4
2 | Onboarding drop 38% (funnel)  | Shorten signup form      | 240
3 | Churned acct cited exports    | CSV export on reports    | 4.0

Want me to create, assign (Ada · Backend AI), and schedule all 3 into Sprint W27?
```

Every number traces to a record on the spine — accounts, MRR, funnel steps — not a planning-poker vibe. Effort stays a human call, and nothing is written until you confirm.

## Other ways to connect

- **Any MCP host:** custom connector URL `https://platform.aioproductos.com/api/mcp`
- **CLI/stdio:** `npx @aioproductoscom/mcp`
- Docs: [aioproductos.com/mcp](https://aioproductos.com/mcp)

## Support

Questions, feedback, feature requests: [office@aioproductos.com](mailto:office@aioproductos.com) — a human founder answers every email.

## License

MIT — see [LICENSE](./LICENSE). The plugin is free; it drives a paid AIOProductOS workspace.

TDQS

A3.9/5.0

Scored across 71 tools

Disambiguation5/5

Despite the large number of tools, each targets a distinct resource and action; reads, writes, analytics, comms, bookings, identity, and artifact-review operations are cleanly separated. Potentially similar pairs like list_ideas/list_insights and pm_meta/list_features are explicitly cross-referenced so an agent can tell them apart.

Naming Consistency4/5

The overwhelming majority follow a consistent verb_noun pattern: list_*, get_*, create_*, update_*, analyze_*, with action verbs like merge, unmerge, vote, promote, and reschedule. Minor deviations like pm_meta and whoami break the pattern slightly, but they are isolated and easy to learn.

Tool Count1/5

71 tools is an extreme number for a single MCP server, well above the 50+ threshold for a severe mismatch. Even though the tools are grouped by domain, the surface is very heavy for an agent to navigate and would be better split into focused servers such as PM, analytics, comms/support, identity, and bookings.

Completeness4/5

The core product-management lifecycle is well covered: features, tasks, sprints, objectives, key results, initiatives, ideas, experiments, releases, decisions, pages, and insights all have list/create/update flows, with archiving and status transitions replacing hard deletes. Minor gaps exist—no booking creation, no conversation reopen/assign, and no delete for most entities—but agents can work around them without dead ends.

Maintenance

ActivitySlowing
ResponsivenessUnresponsive