Skip to main content

​
The workflow

You’ve written some code and want a narrated video walkthrough. With the pull/story MCP server installed, this is one sentence to Claude Code.
1

Make sure the MCP is installed

See the Claude Code setup guide if you haven’t yet.
2

In Claude Code, type

3

Claude Code will

  1. Identify the files that matter
  2. Read tight line ranges from those files
  3. Extract the runtime flow and important gotchas
  4. Call the pull/story MCP tool create_pull_story with a structured brief
  5. Poll get_pull_story until ready
  6. Paste the URL back into the chat
4

You get

A clickable pull story URL in the chat, ready to share with reviewers or embed in Slack.

​
Example conversation

You: Make a pull story explaining the payment retry flow. Claude Code: I’ll inspect the payment retry code path and build a concise brief. [reads src/payments/retry.ts, src/jobs/payment-retry-worker.ts, and src/lib/billing.ts] Generating pull story… [calls create_pull_story({title, summary, repo, files, flow, highlights})] Generating… this usually takes about 90 seconds. Polling… Pull story ready: https://pullstory.com/s/brief_0123abcd4567ef89 Duration: 2m 18s · 7 scenes

​
Variations

You: Walk me through how login sessions work Claude Code finds the relevant auth files, reads them, then calls create_pull_story.
You: Make that pull story again but focus only on retry backoff Claude Code builds a narrower brief and calls create_pull_story again.

​
Why this beats copy-pasting diffs

Traditional flow: paste snippets into chat → explain architecture from memory → answer 3 rounds of clarifying questions. Pull/story flow: say one sentence → get a video link grounded in real files. Teammates watch 2 minutes and actually understand the system before leaving feedback.

​
Next

GitHub App auto-gen

Install once — skip the “ask Claude Code” step entirely.

Embed pull stories

Drop a pull story inline in docs, changelogs, blog posts.