Open source

Free public knowledge bases, edited by AI.

FreeDocStore publishes public docs from GitHub. Ask AI to update a page, review the diff, then apply through GitHub. No CMS body editor, no hidden database, no manual text changes outside source control.

The problem

Public knowledge gets scattered, stale, and hard for agents to trust.

Teams write docs in READMEs, wikis, notes, and chat threads. AI can help maintain them, but only if the source is inspectable, reviewable, and published in a format both humans and agents can read.

Scattered

Knowledge lives in too many places

A public project can have useful docs split across Markdown, GitHub issues, generated pages, and old website folders. Readers and agents need one published surface.

Stale

Docs fall behind the source

Most docs platforms make editing easy but review weak. FreeDocStore keeps GitHub as the source of truth so every change has history, authorship, and rollback.

AI drift

AI edits need review boundaries

FreeDocStore lets AI draft a replacement, but humans review the diff. If a person wants to hand-edit, they do it in GitHub rather than a separate CMS.

Agent-ready

Docs need machine-readable outputs

Public KBs should ship search, sitemaps, `llms.txt`, and MCP discovery metadata so agents can cite and update the right source.

Published KBs

Each knowledge base is a Zensical repo.

FreeDocStore publishes GitHub repositories that use Markdown in docs/ and a zensical.toml config. Each KB deploys to its own Cloudflare Pages project and can attach custom domains.

Draft 0.1

True Non-Profit

A first-principles body of knowledge about non-profits, what they should be, and how to assess their trueness.

Read the KB

Source

GitHub-backed Markdown

The first KB lives in its own repository. Zensical builds it from Markdown, and Cloudflare hosts the generated site.

Open the repo

AI-first editing

No CMS text boxes. No silent rewrites.

The editor is a proposal surface. People describe the change, AI drafts the replacement, and the final manual intervention happens in GitHub where branch protection, review history, and blame stay intact.

Prompt

Ask for the change

Use natural language to update a page, add policy detail, tighten an explanation, or convert notes into publishable docs.

Review

Inspect the proposal

The UI shows the current source, the AI replacement, and a line-level diff. It does not expose a manual content editor.

GitHub

Manual edits stay in GitHub

If a human needs to touch text by hand, the editor opens the GitHub file editor for the target path.

Extension

Edit from published docs

The browser extension uses the same rule from a live docs page: prompt, preview, then Apply through GitHub.

The loop

Prompt the change, review the proposal, publish from GitHub.

The loop is intentionally narrow. FreeDocStore is not trying to replace GitHub; it makes GitHub-backed docs easier for humans and AI agents to maintain.

1

GitHub holds the source

Connect a public Zensical repo, choose the docs folder, and publish through Cloudflare. Source files remain in the repository where review, permissions, and history already exist.

2

AI drafts content changes

Ask for a rewrite, new page, cleanup, or structure pass. The editor creates a proposal and diff, not a hidden mutation.

3

Public output stays cheap

Free public KBs are built by Zensical with generated search, sitemap, metadata, and future MCP discovery.

How it works

Start with a Zensical repo, grow into an agent-readable KB.

  1. 1

    Connect a public repo

    Register the GitHub repository and docs path. FreeDocStore reads source from GitHub instead of storing a second copy.

  2. 2

    Publish the knowledge base

    Build the Zensical project with navigation, search, sitemap, and agent-readable metadata.

  3. 3

    Ask AI for updates

    Use the web workbench or extension to request changes. The AI produces a proposal and diff, then GitHub handles manual review.

  4. 4

    Expose docs to agents

    Publish `llms.txt`, source metadata, and future MCP discovery so agents can find and update the right pages.

docs/auth/session-flow.html
<!-- Written 2026-04-23. Approved by @sergey. -->
<article class="doc">
  <h1>Session flow</h1>

  <p>Sessions are issued as signed cookies.
  Rotation happens on every privileged action.</p>

  <h2>Invariants</h2>
  <ul>
    <li>No token in localStorage.</li>
    <li>Cookie is httpOnly and SameSite=Lax.</li>
  </ul>
</article>
Agent-readable docs

Especially useful when AI is maintaining your knowledge base.

Agents need current source, clear context, and permission boundaries. FreeDocStore gives them public docs to read and a proposal workflow for updates.

Read

Agents get stable public context

Published pages, search data, sitemaps, and future `llms.txt` output give agents a reliable place to retrieve the current explanation.

Propose

AI drafts updates without taking control

The editor gives AI enough source context to propose a full replacement while keeping apply/review in the user's hands.

Trace

GitHub keeps authorship visible

Every accepted update can become a commit or PR. That keeps history, blame, branch protection, and rollback where teams already work.

Free

Public knowledge stays public

The free product is for open/public KBs. Private staff or customer docs belong in ProDocStore.

MCP

Agents can inspect and plan Zensical KB publishing.

The FreeDocStore MCP server exposes the public registry, repo inspection, Zensical validation, deploy checks, and publish planning. Write tools will be added behind auth.

Live

Remote MCP endpoint

Use the current Cloudflare Workers endpoint while the custom MCP domain is being attached.

Open endpoint notes

Discovery

Machine-readable connector

Agents can discover the MCP server from the platform's well-known metadata.

Open MCP discovery

Publish a public knowledge base.

Use the web editor or extension to start from a GitHub-backed docs folder. FreeDocStore is open source, MIT licensed, and free for public docs.