Published: 2026-06-09

Kilo Code + Claude Fable 5: Replicating GitHub From a Single Screenshot

Chapters / key moments (click to jump — plays here on the page)

The Kilo Code team put Claude Fable 5 to the test immediately after its Anthropic release — by pasting a screenshot of the GitHub website and prompting it to recreate the full platform including functional Git repository hosting. After about 10 minutes of autonomous coding, the output was a near-exact visual replica of GitHub with working mock repository features.

Source video

"I replicated GITHUB in one shot with Anthropic Fable 5!" by Kilo Code — Watch on YouTube →

Key Takeaways

  • Fable 5 (Anthropic's latest model, described as a safeguarded version of Mythos) is already available in Kilo Code — select it from the model dropdown; no separate API key needed with Kilo Tokens.
  • The test prompt was simple: paste a GitHub screenshot → ask Fable 5 to "recreate GitHub including the actual functionality of hosting Git repositories."
  • After ~10 minutes of autonomous coding, the result was a one-to-one visual replica of GitHub UI with functional mock Git repository features — no existing codebase read.
  • Demonstrates Fable 5's ability to reconstruct complex web apps from visual screenshots alone — matching Anthropic's claim that Fable can "rebuild a web app's source code from simply reviewing UI screenshots."
  • Kilo Code supports 500+ models including the full Anthropic lineup. Switch to Fable 5 in the model selector mid-project without losing context.

What you can actually set up from this

Extracted from the video's own transcript — the specifics the original summary left out.

Reproducible steps

  1. Select the model

    In Kilo Code's VS Code extension, pick Claude Fable 5 from the model drop-down.

  2. Paste a screenshot and describe the goal

    Paste a screenshot of the GitHub website and ask it to recreate GitHub including the real functionality of hosting Git repositories.

  3. Wait, then test locally

    The run took about 10 minutes and cost $4.07. The local app matched the screenshot closely with mock repositories, and the tester created a new repo and pushed a real codebase to it from the terminal.

  4. Switch models for follow-up edits

    Fable was Anthropic's most expensive model at the time, so for further edits switch to a cheaper one such as Sonnet 4.6; Kilo lists 500-plus models.

Gotchas

  • A single vendor-run test on a model launched that week; the $4.07 and 10 minutes are one run and will vary with the prompt.

Weekly Digest — In Your Inbox

Get the week's top AI agent news, updates, and guides — every Friday.