Skip to content

Public Tools

Free AI Website Legibility Tools

Audit how search engines and AI agents read your website. Generate llms.txt, check robots and structured data, compare visible copy with metadata, and inspect agent activity.

Tools and experiments

Public tools, built to be used

These free tools help website owners and developers see what search engines and AI agents can read. They expose missing signals, generate useful files, and compare public claims across pages and metadata. Other tools make code and agent activity easier to inspect and verify.

The web legibility suite forms one loop. Audit with Legible to find what a model cannot read. Generate the llms.txt it flags as missing. Then use Aligned to verify that visible copy agrees with metadata, structured data, and AI-facing files. Every result comes from what your site publishes.

The web legibility loop

Developer tools & experiments

MarginThe modern Markdown editor for you.A Markdown editor for poets, storytellers, and anyone who writes. Draft in a calm serif with syllables, rhyme and stress in the margin, or switch to source view with code colours and eleven themes. No account, and your words stay on your device until you export them.Start writing Agent FoundryGive your agent a purpose.Describe what an agent should do in a sentence and get editable instructions for ChatGPT and Claude projects, Codex, Claude Code and its subagents, OpenClaw, Agent Skills, and Microsoft Foundry. Shape them in a real editor, then copy one or download all seven with setup guides. Everything is generated in your browser, without an account or an inference service.Open the foundry izakayaYour repos, served as small plates.A zero-dependency terminal app that turns your code directory into a working menu: git status, recent commits, diffs, activity, languages, stack, stashes, unpushed work, and what changed since your last visit. Read-only, fast, and open source under MIT.Pull up a stool ChainProofA verifiable record of every AI action.An API-first provenance ledger for AI agents. It appends actions, decisions, approvals, and artifacts to a SHA-256 hash chain, making later edits detectable and independently verifiable. The MIT-licensed app makes one deliberately narrow guarantee.Explore ChainProof EigenstateA screensaver with a memory.Agents reorganize, experiments change, and a rotating 3D world responds. Close the tab and come back later, and your universe is still there. Everything is simulated; the computation and saved state stay in your browser. Open it, go fullscreen, and press W to let the world model take over.Open Eigenstate

izakaya and ChainProof bring the same approach to developer tooling. They make hidden state visible and let people verify the result. Both are open source under the MIT license.

I built Eigenstate with GPT because I liked the idea of a screensaver with a memory, with the parts connected so a change in one system leaves a trace in another. I wanted the movement to have a reason, even if you never look closely enough to work out what it is.

Built in the open, with outputs you can inspect.

Start a Conversation
What brings you here?

Your information is used only to respond to this inquiry.

Prefer another channel? Find Matthew on LinkedIn or explore his work at Clevyr.