All posts
Claude Code for product managers: a practical guide
How PMs use Claude Code: where to run it without a terminal, how to get started, the workflows that work, and how to take changes to your team's repo with engineers approving.
Ben Ong
October 30, 2026
product-development
ai-agents
A PM request flowing to a plan that three teammates join, then to a live preview, then to a merge with a check mark, on a textured orange background
Can product managers use Claude Code? Yes. Claude Code is Anthropic's AI coding agent, and you don't need to be an engineer or live in a terminal to use it. It also runs in a desktop app, in the browser and inside code editors. PMs use it to ask questions about their product's code, prototype changes on the real codebase, and pull context from tools like Jira or Linear. The harder part is getting a change from your session into the product your team ships, with engineers reviewing it.
This guide covers where to run Claude Code, how to get started, the PM workflows that work well, and how to take changes to your team's repo.
What Claude Code is, in PM terms
Anthropic's overview describes Claude Code as "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools." In practice, you describe what you want in plain language. Claude Code finds the relevant files, proposes a plan, makes the changes and can open a pull request.
For a PM, that means three things:
- You can ask the code directly. "Does the onboarding flow already support SSO?" gets an answer from the source, not from memory.
- You can prototype on the real product. A change built on your actual codebase shows what the feature would really look like.
- You can connect your other tools. Through MCP (Model Context Protocol), Claude Code can read from and act in tools like Linear, Jira, Notion or Google Drive.
Where you can run Claude Code
Many guides start with the command line, but it's only one option. According to Anthropic's platforms page, Claude Code runs on these surfaces:
| Surface | What it's like | Good fit for PMs? |
|---|---|---|
| Terminal (CLI) | The full-featured version, run from a command line on your machine | If you're comfortable with a terminal |
| Desktop app | A standalone app with a visual diff viewer and app preview | Yes, if you want a local setup without the terminal |
| Web (claude.ai/code) | Cloud sessions in your browser, with no local setup | Yes, for repos connected through GitHub |
| VS Code and JetBrains | An extension or plugin inside your code editor (the JetBrains plugin also needs the CLI installed) | If you already use an editor |
| Mobile | The Code tab in the Claude app, for starting and checking on cloud sessions | Best for starting and checking on tasks away from your desk |
Anthropic's quickstart lists the accounts you can sign in with: a Claude Pro, Max, Team or Enterprise subscription, a Claude Console account, or a supported cloud provider. Cloud sessions on the web need a Pro, Max or Team plan, or an Enterprise premium or Chat + Claude Code seat.
How to get started with Claude Code as a PM
- Pick a surface. If the terminal feels like a barrier, start with the desktop app or Claude Code on the web.
- Get access to the code. Ask engineering which repository you can use and whether you should work on a branch. For cloud sessions, Claude Code needs a GitHub connection to clone the repo and push branches.
- Start by asking questions. Try "what does this project do?" or "where is the checkout flow handled?" before you change anything. It's the fastest way to learn the codebase.
- Use plan mode before edits. In plan mode, Claude reads files and proposes a plan without editing anything until you approve. It's a good habit when you're not sure what a change will touch.
- Give it standing context. A
CLAUDE.mdfile holds instructions Claude reads at the start of every session, such as how the project is structured or which conventions to follow. Anthropic's CLAUDE.md docs note these files are shared with the team through version control, so agree on them with engineering. - Connect the tools you use every day. Add MCP servers for your tracker and docs so Claude can pull tickets, PRDs and specs into the session.
Ways product managers use Claude Code
The most useful PM guides focus on everyday product work, not on writing code for its own sake. The workflows that come up most often:
- Checking whether a feature exists. The PM Skills practical guide shows asking Claude in plain English whether a feature is in the code, and checking whether what's "Done" in Jira is actually done in code.
- Prototyping on the existing codebase. The same guide builds a quick UI prototype on top of a real app to test an idea and make a PRD more concrete.
- Summarizing sprints and backlogs. With a Jira or Linear MCP connected, you can ask why a ticket is blocked or whether the backlog has duplicates.
- Writing and sharing documents. In Aakash Gupta's advanced guide, Carl Vellotti sets up MCPs for Notion and Google Drive to get documents in and out of Claude Code.
- Turning repeat work into commands. Both guides recommend packaging repeat work as skills or commands, so the next run is one step. The PM Skills example is a /standup shortcut.
Where solo Claude Code gets harder for PMs
Claude Code works well when you're the only person who needs to see the result. Shipping a product change usually involves other people, and that raises questions a single session doesn't answer on its own:
- Who reviews the change? On many teams, an engineer approves a pull request before it merges. Your change still has to go through that review.
- Can others try it? A designer or stakeholder may want to click through the working change before anyone reviews the code.
- Who else is in the loop? If a designer wants to adjust the plan or an engineer wants to steer, they need a way into the work while it's in progress.
- What can you access? Engineering may want limits on which repositories and branches non-engineers can change.
None of this is a flaw in Claude Code. It's how teams ship safely. But it's where a PM's solo prototype often stalls.
How to do this with your team in LightSprint
LightSprint is a separate product from Claude Code. It's the shared workspace for teams that want PMs, designers and ops to make real changes on the existing codebase, with engineering approving every merge.
- Use Claude or Codex in a shared session. There's no local setup. Connect your existing Claude or Codex subscription, or use your own model API key.
- Start from a visual plan. Describe the change in plain language. LightSprint drafts a plan from your existing codebase, and the team can review it before any code changes.
- Work together in the same session. Anyone on the team can join the session and guide the agent while it builds.
- Click through a live preview. Each change runs in an isolated sandbox with a live preview, so PMs, designers and stakeholders can click through each change.
- Hand engineering a pull request. As a Builder, you can propose changes and view previews but you can't merge. Engineers approve the PR through the branch rules and required reviews your team already uses.
There's more on the PM workflow on our product managers page, and on access controls on our security page. If you're a developer comparing the tools, read LightSprint vs Claude Code.
FAQ
Can product managers use Claude Code without a terminal?
Yes. Anthropic offers Claude Code in a desktop app, on the web at claude.ai/code and in VS Code and JetBrains, as well as in the terminal. The web version runs in the cloud with no local setup, for repositories connected through GitHub.
Do I need to know how to code to use Claude Code?
You don't need to write code, because you describe changes in plain language. It helps to understand what a change does to the product, and plan mode lets you review the approach before anything is edited. Engineers should still review code before it ships.
Do I need a paid plan for Claude Code?
According to Anthropic's quickstart, you sign in with a Claude Pro, Max, Team or Enterprise subscription, a Claude Console account, or a supported cloud provider. Check Anthropic's site for current plans.
Is it safe for a PM to use Claude Code on the company codebase?
Agree on the ground rules with engineering first: which repository, which branch, and how changes get reviewed. Plan mode and permission settings let you check what Claude will do before it does it.
How is LightSprint different from Claude Code?
Claude Code is a coding agent. LightSprint is a shared workspace where your team uses agents like Claude and Codex together, with shared sessions, live sandbox previews and engineering approval before anything merges. The full comparison is in LightSprint vs Claude Code.