Changelog
Last updated: May 2026
All notable changes to Contxt will be documented here.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[Unreleased]
Added
- Brand guidelines page (
brand/guidelines.html) - Privacy Policy and Terms of Service pages (
/privacy,/terms) - Contributing guide (
CONTRIBUTING.md,/contributing) - Changelog page (
CHANGELOG.md,/changelog)
Planned
- User dashboard with link management
- Link analytics (views, referrers, devices)
- Link expiration and password protection
- Public API with API keys for programmatic creation
- Plan enforcement (Free / Pro tiers)
- Team / collaboration features
- MCP server for direct AI tool integration
[0.1.0] — 2026-05-14
Added
Core Experience
- Paste-to-link flow: enter content → AI generates title + summary → short link created
- Guest creation — no account needed to share context
- Optional email claim to receive visit notifications
- Slug generation: 2-word prefix + 5-char random code with collision retry
- QR code generation for every shared link
- Copy / Open / QR actions on the result card
Landing Page
- Full hero section with tool card (paste, create, preview in one card)
- "How It Works" three-step explainer
- "The Experience" section with dual browser/AI YAML views
- Use Cases section with hover cards
- Scroll-triggered fade-up animations
- Brand: warm beige/pink palette, Playfair Display + DM Sans typography
Shared Context Pages (/s/[slug])
- Browser view: rich card layout with title, summary, full markdown content
- AI agent view: structured YAML frontmatter served as
text/plainvia proxy rewrite - Both views provide "Continue in ChatGPT / Gemini / Claude" action buttons
- Fallback prompt option for AI agents that can't POST to the API
- Smart preview cards for Slack, email, and doc embeds
AI Agent Support
- User-agent-based routing (browser vs ai_agent)
- Treats curl and wget UAs as AI agents for debugging
- Explicit YAML format: name, description, content fields
- Server-side proxy rewrites AI agent pages to Content-Type: text/plain
- Agent landing page (
/api/ai/page) — structured API reference with examples - Agent context endpoint (
/api/ai/context/[slug]) — JSON context data - Fallback HTML-to-text extraction for agents that can't POST
API & Backend
POST /api/contexts— create a context (JSON body)POST /api/generate-meta— AI-generated title + summaryPOST /api/claim— claim a link with emailGET /api/qr— QR code generation- Server action
createContext(guest flow, no API needed) - Input validation via Zod schemas
- Turso (libSQL) database with Prisma ORM
- Lazy
getPrismaClient()for Vercel serverless compatibility
Fixed
- Dynamic origin for "try this prompt" URLs (was hardcoded to mvp.contxt.to)
- URL generation uses actual request host (not hardcoded domain)
- Result card uses API-returned URL
DB2_TURSOenv fallback for preview database- Clean AI YAML — removed tags, slug, timestamps; kept only name + description + content
- Markdown content is properly encoded for AI agent pages
Changed
- Landing page redesigned from concept mockup to production v7 (warm beige/pink)
- Browser context page redesigned matching v2 design spec
- AI agent pages: YAML frontmatter → cleaner format
- Form simplified: one content field, AI handles title + summary generation
Removed
- Tags, slug, createdAt, claimToken from AI YAML output
- Dashboard concept mockup section from landing page
- Unused slug/tags/createdAt fields from ContextView browser page
Security
- Input validation on all API endpoints (Zod)
- IP-based rate limiting structure in place
- Content max length: 50,000 characters
v0.0.x (Pre-release)
- Dashboard concept mockup iterations (added, reverted, reapplied, refactored)
- Initial Vercel SPA configuration with /api/* exclusion for serverless functions
- AirTable integration exploration (deprecated)
- Initial project scaffolding and MVP foundation