data-liberation-agent

Liberate any website into a complete, portable HTML site.

The problem

Closed platforms make it hard to leave. Wix has no HTML export and caps RSS at 20 posts. JavaScript-rendered content and limited APIs leave your site locked inside.

The solution

Point it at a URL and get your site back as plain files:

data-liberation https://example.com/

Every retained route becomes a directory of HTML, CSS, media, and fonts, with references rewritten so navigation works locally. It runs on its own, anywhere that can serve a folder.

HTML is the contract. The liberated site is the deliverable — not an intermediate format on the way to somewhere else, and not tied to any destination.

Four commands

data-liberation <url>                        # liberate a site
data-liberation inspect <url>                # assess a source before liberation
data-liberation compare <run-dir>            # verify the copy against its source
data-liberation publish <run-dir> --to spacefast   # put it on a live URL

That is the whole surface. Inspection is bounded and read-only; it reports measured route/document facts, coverage, issues, and explicit unknowns without writing a site. Liberation writes the site and exits; add --serve to keep a local server running so you can click through it.

See rendered inspection and complexity and source cleanup, comparison, and destination attribution for their API contracts, evidence, and limits.

Embedded Node consumers can import the same operations from data-liberation/runtime, or use the committed standalone dist/capture-engine.bundle.mjs. Both expose inspect, capture, compare, publish, and their extension registries.

Inspect

data-liberation inspect https://example.com/ --discovery-limit 50 --sample-limit 5

The JSON result is versioned and destination-neutral. It samples the entry route plus distinct discovered route types, up to the configured limits. It does not claim browser-rendered layout, responsive reflow, or activated interactions: those remain explicit unknowns until full liberation and comparison.

Liberate

Flag
--output <dir>	Output base. Default ~/data-liberation, or DLA_OUTPUT_DIR
--resume	Reuse what is already on disk instead of recapturing
--screenshots	Also write full-page desktop and mobile PNGs
--serve	Keep a local server running until interrupted
--no-learn-fluid	Freeze the layout at one width instead of learning how it reflows

Fluid learning is on by default. The capture sweeps widths with the source's own runtime alive, fits how each element is sized, and emits that as ordinary CSS — so the copy keeps reflowing after the runtime is stripped, instead of being pinned to the width it was captured at.

Compare

Verification runs in two tiers, and the report says which found what.

• Self-consistency, across every route, offline: anchors resolving to exactly one target, internal links landing on a real file, no asset still pointing at the origin. Milliseconds.
• Source fidelity, across a sample, in a browser: text, geometry and reflow at widths the capture never sampled, and dialogs opening as the source opens them. Roughly 25 seconds per route, so it samples rather than walking everything.

Exit code 0 means both passed. --screenshots writes source, copy, and diff PNGs; the pixel score is evidence for a human and never decides pass or fail.

Publish

data-liberation publish <run-dir> --to spacefast

Publishing reads what is on disk, so a repaired copy or a different target can ship without touching the source site again. Without a token the publish is anonymous and returns a one-time claim link — the space stays private until it is claimed, so the live URL answers 403 to everyone until then. --token, or SPACEFAST_TOKEN, publishes into an account you own.

Naming an unknown target lists the registered ones.

Supported platforms

Platform	Status
GoDaddy Websites & Marketing	Ready
Hostinger Website Builder	Ready
HubSpot	Ready
Shopify	Ready
Squarespace	Ready
Webflow	Ready
Weebly	Ready
Wix	Ready
Any other website	Best-effort generic fallback

Adapters contribute platform knowledge to discovery and capture — how a platform lists its routes, what its CDN URLs look like, how its runtime resolves anchors. Sites matching none of them fall back to a generic adapter that renders each page in a headless browser.

Custom platforms

Consumers can register a platform that auto-detects, discovers routes, and applies source-specific liberation hooks without editing core. See Platform API for the public contract and example.

Output

A run produces, under ~/data-liberation/<host>/:

Path
website/	The deliverable. Serve this directory anywhere
capture-receipt.json	Source URL, the route table mapping each page to the URL it came from, assets
asset-evidence.json	Bounded v1 source-asset evidence. Each source URL records retained-route references, retrieval separately from portable inclusion, and its canonical portableAssetId when bytes were deduplicated; consumers group source records by that ID. It retains the first 10,000 reachable source URLs in captured-route traversal and emits them lexically; assetCount is exact unless assetCountExact is false, when it is a lower bound and assetsTruncated identifies omitted detail. totalReferenceCount remains exact, while per-asset referencesTruncated identifies omitted locations. The receipt links this sidecar as assetEvidence.path.
diagnostics.json	Everything the source withheld or the capture could not resolve
source-profile.json	Measured behaviour: one document or per-device, declarative or runtime-written geometry, switch width

Using it from an agent

The liberate skill drives the CLI. It is the only skill: a URL becomes a verified, portable site, and optionally a live URL.

Claude Code

claude plugin marketplace add Automattic/data-liberation-agent
claude plugin install data-liberation@data-liberation

Then /liberate https://your-site.com.

Codex

cd data-liberation-agent && codex

.codex-plugin/plugin.json and .mcp.codex.json register the server and skill automatically, since Codex does not expand ${CLAUDE_PLUGIN_ROOT}. Then $liberate https://your-site.com.

Gemini CLI

cd data-liberation-agent && gemini extension link .

Any MCP client

npm run mcp     # or: npx tsx src/mcp-server.ts

The server exposes the same four verbs as tools — inspect, liberate, compare, publish — calling the same entry points the CLI calls. MCP is a transport here, not the architecture: nothing requires it, and it deliberately does not expose internal pipeline phases.

First-time browser setup. Capture uses Playwright's Chromium, which is not installed automatically:

npm run setup:browser

Additional documentation

• Platform API — custom platforms
• Wix authenticated content endpoints — the load-bearing content endpoints behind Wix's editor and dashboard auth
• Discoveries — findings from real runs

Related

• WordPress Data Liberation project — the official effort
• Spacefast — agent-first static hosting, the default publish target