Product teams are starting to use AI in two different ways.

They ask AI to help them understand what is going on.

And they ask AI to help them build.

Those two workflows should not be separate from customer feedback.

If a designer is exploring a problem, the answer should be grounded in what customers said. If a PM is comparing roadmap bets, the recommendation should include evidence. If an engineer is planning an implementation, the agent should understand the customer pain behind the work. If a founder or product leader is reviewing priorities, the discussion should not drift away from the source material.

That is the idea behind Deck plugins and skills for Claude Code and Codex.

The goal is not only to make coding agents more informed.

The goal is to make customer evidence available wherever product teams are thinking, planning, deciding, and building with AI.

Two use cases, same evidenceCopy heading link

There are two use cases we care about equally.

The first is finding useful customer information.

Product teams should be able to ask Claude Chat, Claude Cowork, Claude Code, Codex, or another AI workspace questions like:

• "What customer problems are getting worse?"
• "Which roadmap opportunities have the strongest evidence?"
• "What should we prioritize for onboarding?"
• "What are designers hearing about this workflow?"
• "Why are enterprise customers churning?"
• "What are detractors saying in NPS?"

The second is building with customer context.

When a team moves from evidence to action, the agent should keep the customer context in the work:

• drafting an initiative brief
• preparing an engineering handoff
• checking whether a planned change matches the original customer problem
• reviewing implementation tradeoffs against customer pain
• writing launch notes from the evidence behind the feature
• identifying missing research or weak assumptions before the team commits

Both workflows need the same foundation: Deck customer evidence.

The problem is not that teams lack feedbackCopy heading link

Most teams already have feedback.

They have interview recordings, support tickets, Slack messages, survey responses, NPS comments, sales notes, CSVs, research docs, and meeting notes.

The problem is that feedback is difficult to use at the moment product work happens.

When a PM is deciding what to prioritize, they should not have to manually open five tools and turn scattered notes into a defensible recommendation.

When a designer is shaping a flow, they should not have to guess whether the pain is broad, recent, segment-specific, or just loud.

When an engineer is asked to build something, they should not have to rely only on a ticket summary if deeper customer context exists.

When a leader asks "what should we bet on?", the answer should not come only from instinct, stakeholder pressure, or the loudest recent conversation.

Deck already turns scattered feedback into themes, insights, quotes, opportunities, initiatives, segments, and NPS analysis.

The plugin and skills bring that evidence into the AI tools product teams already use.

What the Deck skills help withCopy heading link

The Deck plugin bundles five skills designed around recurring product workflows:

These are not meant to be novelty prompts.

They are repeatable product workflows.

For discovery and prioritization, you might ask:

Use Deck to compare the top three roadmap opportunities for activation.
Rank them by evidence strength, customer reach, sentiment, recency, and risk.
Include representative quotes and open questions.

For design or research, you might ask:

Use Deck to find the customer problems with the worst sentiment this month.
Group them by theme, affected segment, severity, and whether there is already an active initiative.

For planning and execution, you might ask:

Use Deck to create an initiative brief for improving onboarding.
Include the customer problem, target segment, evidence, proposed scope, non-goals, success metrics, risks, and next step.

For engineering handoff, you might ask:

Use Deck to summarize the customer evidence behind this feature area.
Then compare it with the current implementation plan and call out where the plan does or does not address the customer problem.

Installed workflows and portable workflowsCopy heading link

There are two ways we want product teams to use these workflows.

The installed workflow is the plugin.

Claude Code and Codex can connect to Deck through MCP, authenticate with browser OAuth, and query Deck directly. That is useful when teams are planning, implementing, reviewing, or working close to the codebase.

The portable workflow is the skills page we are adding to the Deck marketing site.

That page will show the Deck skills in a copy-paste format, so product teams can use the same workflows in Claude Chat, Claude Cowork, or another AI workspace where they already do product thinking.

The plugin is for connected agents.

The skills page is for copy-paste workflows.

Both are meant to do the same thing: make it easier for product teams to use customer evidence at the moment they need it.

What is includedCopy heading link

The plugin includes Deck's hosted MCP server:

https://mcp.getdeck.io/mcp

Deck MCP uses browser OAuth. Users do not need API keys.

An org admin must enable MCP access in Deck from Settings -> MCP before tools can read feedback.

Install for Claude CodeCopy heading link

Add the marketplace:

claude plugin marketplace add getdeck-io/plugins

Install the Deck plugin:

claude plugin install deck@deck-plugins

After installing, start Claude Code and run:

/mcp

Authenticate with your Deck account when prompted.

To confirm the server from the shell, run:

claude mcp list

If you need to add the server manually, run:

claude mcp add --transport http deck https://mcp.getdeck.io/mcp

Install for CodexCopy heading link

Add the marketplace:

codex plugin marketplace add getdeck-io/plugins

Install the Deck plugin:

codex plugin add deck@deck-plugins

Confirm the server is installed:

codex mcp list

Authenticate through the browser OAuth flow:

codex mcp login deck

If you need to add the server manually, run:

codex mcp add deck --url https://mcp.getdeck.io/mcp

The shiftCopy heading link

AI should not make product teams less customer-led.

It should make customer evidence easier to use when teams are asking questions and when they are building.

That means helping designers, PMs, engineers, founders, and product leaders find useful customer signal and carry that context into the work.

That is what Deck plugins and skills are for.

You can find the plugin repo at github.com/getdeck-io/plugins.