DocsAI — Product Requirements Document

public docs-ai v2 ·by andre ·Mar 22, 2026 productoreference

One-liner

AI-native client deliverable portal for agencies. AI agents publish markdown → clients see branded, access-controlled pages with view tracking. Open-source engine + hosted cloud (Vercel/Cal.com model).

The Problem

Every company using AI agents (Claude Code, Cursor, Copilot) generates markdown files that sit in git repos, invisible to anyone who doesn’t cat files in a terminal. The output is trapped.

The gap between “AI generated a great document” and “client can see it professionally” is manual work: copy to Notion, export to PDF, upload to Drive, send link. That manual step kills the speed advantage of AI.

For agencies specifically: multiple clients × multiple deliverables × need for branded, access-controlled output = an unsolved workflow.

The Insight

Documents shouldn’t be edited by humans anymore. They should be authored by AI agents and published automatically. The human’s job is to review, approve, and share — not to format, upload, and manage links.

The defensible version isn’t “we render markdown better” — it’s “the only platform purpose-built for agencies to publish AI-generated deliverables for clients, with workflow depth that documentation platforms won’t build.”

Analogy: DocSend meets GitBook, but for AI-native agencies. Content stays in git (zero vendor lock-in); the platform handles branding, access control, view tracking, agent-to-agent workflows, and the “last mile” to non-technical stakeholders.

Users & Personas

Persona	How they interact	Primary need
AI Agent (Claude Code, Cursor, Codex, ChatGPT)	API / MCP / CLI skill (`/publish`)	Push markdown → get URL back. Discover capability via `llms.txt` + `.well-known/*`.
Team Member (agency staff)	Web UI, light editing	Browse, organize, approve docs.
Client (external)	Read-only branded view	See deliverables professionally. Approve, comment, download.
Public visitor	Blog / case study pages	SEO-friendly content, brand presence.

Priority order: AI Agent > Client reader > Team member > Public visitor.

Core Concepts

Document

A markdown (or MDX) file in a project folder with YAML frontmatter:

---
title: "Brand Guidelines — Acme Corp"
role: client
version: 3
author: andre
created: 2026-03-22
tags: [marketing, reference]
---

Body uses controlled tag taxonomy (Domain × Type, optional Subdomain) and can compose components from @cofoundy/ui when published as .mdx.

Project Space

Documents grouped by folder. Maps to URL: docs.cofoundy.dev/{role}/{project}/{slug}.

Access Roles

public — open to web
team — agency staff (auth required)
client — only the assigned client (per-project access)

What’s shipped (V1, March 2026)

Astro 6 site with full markdown rendering, dark/light theme, branded typography
Auto-deploy via GitHub Actions on push
PDF document viewer (pdf.js — toolbar, fullscreen, keyboard nav)
Cloudflare Access + D1 middleware for two-layer auth (authentication + per-project authorization)
View analytics dashboard (D1 + tracking beacon)
/publish skill — autonomous e2e: publish + grant access + email notify
26+ documents published across multiple projects

What’s coming next

V1.1 — Agent-Native Surface

Comply with emerging agent-discovery standards:

/llms.txt (root + scoped per-project) — llmstxt.org spec
/openapi.json — full OpenAPI 3.1 for any agent
?mode=agent — clean markdown + frontmatter response on any URL
.well-known/agent.json, ai-plugin.json, mcp.json
MCP server at mcp.docs.cofoundy.com
Hosted skill: npx @cofoundy/docs install
Structured error contract (code, retry_after, docs_url)

V2.0 — Stack Migration

Migrate to Next.js 15+ (App Router, RSC) to consume @cofoundy/ui design system natively. Unblocks visual richness, MDX components, and multi-tenant V3.

V2.1 — Visual Vocabulary

MDX docs compose components from @cofoundy/ui:

<PersonalNote> — signed intro from author
<MetadataCard> — branded chips for “Para / Tiempo / Costo”
<Callout> — note/tip/warn/danger
<ScopeList> — checkmarked deliverable list
<StatCard>, <FunnelChart>, <Heatmap>, <Leaderboard> — embedded analytics in docs
<ChatAboutDoc> — AI Q&A grounded in the current doc
<ApprovalBlock> — signed-off CTA

V2.2 — AI Capabilities Beyond Authoring

URL-driven AI features on every doc:

?lang=es — translate on demand
?explain=section-X — re-explain in simpler terms
?mode=executive — TL;DR summary
Auto-tags + auto-summary at publish
Suggested improvements based on analytics

V2 — Workflow Loop

The actual moat:

Comments + approval with webhook back to publishing agent
Visual version diff (Google Docs “suggesting” mode, not git diff)
Agent-to-agent handoff (drafter → specialist → publisher)
Re-read frequency tracking (“client read your proposal 3 times before signing”)
Section-level analytics (heatmap, scroll depth, time per section)
AI vs human read differentiation
Approval webhooks to CRMs

V3 — Distribution & Open Source

OSS engine + hosted cloud (Vercel/Cal.com/Sentry/PostHog model):

@cofoundy/docs-engine (MIT/Apache) — core open source
Cofoundy Cloud — hosted multi-tenant SaaS
Anonymous publish + claim flow
Custom domains per workspace
Pricing: OSS (free self-host) → Free Cloud → Pro $29 → Agency $99 → Enterprise

Architecture (target state)

Agents (Claude Code / Cursor / Codex / ChatGPT / Hermes)
   │
   ├──/publish skill, MCP, REST API
   │
   ▼
Next.js 15 (App Router, RSC, MDX)
   │
   ├── @cofoundy/ui (design system: 70+ components)
   ├── Agent-native surface: llms.txt, openapi.json, .well-known/*
   ├── AI endpoints: translate, explain, summary, chat
   │
   ▼
Cloudflare Pages + D1 (permissions, analytics, comments, approvals) + R2 (assets)
   │
   └── Cloudflare Access (OTP, OAuth, SSO V3)

Why this could be a startup

Timing — AI coding agent adoption growing 25-28% CAGR. The volume of AI-generated markdown is exploding.
Wedge — Agencies (Cofoundy = customer zero). No markdown-native tool serves this workflow today.
Counter-positioning — Content stays in git, not proprietary blocks. Zero vendor lock-in.
Moat is workflow — Per-client spaces, view tracking, approval flows, agent-to-agent handoff, white-label.
OSS distribution — Differentiates from closed competitors (GitBook, Mintlify), builds community, defangs platform risk.
Expansion — Agency deliverables → consulting → legal (contracts) → marketing (content) → any team using AI for external stakeholders.

Status

V1 shipped 2026-03-25
V1.1 + V2.0 committed (May 2026)
V2.1 / V2.2 / V3 — design phase

For the team-facing extended version (architecture details, risks, open questions, strategic decisions log), see /team/docs-ai/prd-full.