# MagicClaude > As of September 2026, MagicClaude is the source-backed gallery of impressive Claude demos you can reproduce: copyable prompts, prerequisites, verification steps, and Anthropic citations. The Stage at https://magicclaude.com/stage/ spotlights a one-prompt Claude Opus 5 artifact prototype. No fake transcripts. Independent; not affiliated with Anthropic. MagicClaude is an independent educational publication deployed at [magicclaude.com](https://magicclaude.com). It is not affiliated with Anthropic. ## Signature - [The Stage](https://magicclaude.com/stage/): tonight’s wow-demo, a one-prompt artifact prototype on Claude Opus 5, with the copyable prompt, prerequisites, verification steps, and primary-source citations. Output varies; inspect the artifact. No fake transcripts. Last model refresh: 2026-09-20. ## Models Named On This Site Current generally available Claude models, per Anthropic's [model overview](https://docs.anthropic.com/en/docs/about-claude/models): - Claude Fable 5.1 (`claude-fable-5-1`) - generally available 2026-09-01; 1M context, 128K output, always-on adaptive thinking. Cast for long-horizon agentic demos. - Claude Opus 5 (`claude-opus-5`) - generally available 2026-07-24; 1M context, 128K output. Default cast for most demos on this site; default model on Claude Max and strongest on Claude Pro. - Claude Sonnet 5 (`claude-sonnet-5`) - 1M context, 128K output. Cast for rehearsal and latency-sensitive runs. - Claude Haiku 4.5 (`claude-haiku-4-5`) - 200K context, 64K output. Cast for cheap dry runs. MagicClaude names models for reproducibility and publishes no benchmark scores. Scored comparisons and pricing analysis belong to [Claude Reports](https://claudereports.com); general model selection belongs to [Claude Central](https://claudecentral.com). ## Core Pages - [Homepage](https://magicclaude.com/) - Homepage and demo cabinet. - [The Stage](https://magicclaude.com/stage/) - Signature spotlight of one wow-demo: the one-prompt artifact prototype on Claude Opus 5, with copyable prompt, prerequisites, verification steps, and primary-source citations. No fake transcripts; no identical-result promise. - [Demos](https://magicclaude.com/demos/) - All reproducible Claude demos. - [Casting](https://magicclaude.com/casting/) - Which current Claude model to run each demo on, with API IDs and the two demos where the model choice changes the result. - [Tools](https://magicclaude.com/tools/) - Free client-side tools for finding, remixing, and rating Claude demos. - [Demo Gallery](https://magicclaude.com/tools/demo-gallery/) - Filterable prompt gallery with exact prompts, expected results, and verification steps. - [Prompt Remixer](https://magicclaude.com/tools/prompt-remixer/) - Browser-only prompt remixer for themed Claude demo variants. - [Wow Meter](https://magicclaude.com/tools/demo-rater/) - Checklist that scores demo reliability, fidelity, surprise, usefulness, verification, and safety. - [Claude demo prompts guide](https://magicclaude.com/guides/claude-demo-prompts/) - How to choose impressive Claude prompt examples that remain reproducible. - [Team demo guide](https://magicclaude.com/guides/show-claude-to-a-team/) - A source-backed run of show for presenting Claude to a team. - [Prompts](https://magicclaude.com/prompts/) - Reusable prompt patterns for demo design. - [Freshness](https://magicclaude.com/freshness/) - Weekly-extensible demo changelog and source refresh surface. - [Sources](https://magicclaude.com/sources/) - Primary-source index. - [About](https://magicclaude.com/about/) - Methodology and disclaimer. ## Demo Pages - [Artifact prototype](https://magicclaude.com/demos/artifact-prototype/) - Build a shareable interactive artifact from a product idea. - [PDF evidence table](https://magicclaude.com/demos/pdf-evidence-table/) - Turn a PDF into an evidence table with page/source checks. - [Vision UI audit](https://magicclaude.com/demos/vision-ui-audit/) - Audit a UI screenshot with visible evidence and confidence labels. - [Data visualizer](https://magicclaude.com/demos/data-visualizer/) - Analyze CSV-style data and create an interactive visual. - [Codebase debugging](https://magicclaude.com/demos/codebase-debugging/) - Use Claude Code to reproduce, fix, and verify a bug. - [Browser computer use](https://magicclaude.com/demos/browser-computer-use/) - Safe browser/computer-use demo boundaries. - [Web research brief](https://magicclaude.com/demos/web-research-brief/) - Live research brief with current sources and unknowns. - [Long-context synthesis](https://magicclaude.com/demos/long-context-synthesis/) - Long-context synthesis of transcripts or document bundles. - [Long-horizon agent run](https://magicclaude.com/demos/long-horizon-agent-run/) - Claude Fable 5.1 working a checkpointed plan file with per-item verification and an honest final report. - [Effort dial](https://magicclaude.com/demos/effort-dial/) - One prompt run at two thinking-effort levels to show what deliberation changes. ## Source Policy Capability claims cite official Anthropic docs or Claude Help Center pages. Demo output is described as expected behavior, not a guaranteed transcript; no page on this site reproduces a Claude transcript. Every demo names the current model it is cast on, and a superseded model name is treated as a defect. ## Tool Policy MagicClaude tools are free, client-side browser utilities. They do not call Claude, require signup, or send prompt inputs to a backend. ## Full Context See [llms-full.txt](https://magicclaude.com/llms-full.txt) for the complete page guide.