Dump

A collaborative context board for sharing projects between ChatGPT, Claude, coding agents, and humans.

Dump is a visual canvas for collecting links, notes, checklists, and project briefs. Use it as a shared project context layer: one board your team can read, your AI tools can consume, and your future self can actually find again.

Live at dump.page

What Dump Is Good For

• Share a project between ChatGPT and Claude without rebuilding the same context in every chat
• Create a reusable project brief for Claude Code, Codex, Cursor, or other coding agents
• Replace scattered Slack threads and bookmark folders with one shared research board
• Keep multimodal context outside any single LLM vendor or chat history
• Expose board content through MCP and llms.txt for agent-friendly workflows

Features

• Freeform canvas with drag-and-drop
• Link cards with automatic metadata extraction (title, favicon, preview image)
• Text notes with rich text editing
• Checklists
• Board sharing (private, shared via link, or public)
• Real-time collaboration via Convex
• MCP server for ChatGPT, Claude, Codex, Cursor, and other AI tool integrations
• llms.txt endpoints for shared and public boards
• Local-first mode — use without an account

Positioning Angles

Different people find Dump through different problems. These are all valid ways to describe the product:

• Shared project context for ChatGPT and Claude
• AI project handoff board
• MCP-ready context layer for agents
• Research hub for modern product and engineering teams
• Human-and-AI-readable whiteboard for links, notes, and decisions

Tech Stack

• Framework: Next.js (App Router) + React 19
• Backend: Convex (serverless database + functions)
• Canvas: React Flow (@xyflow/react)
• Auth: Firebase (Google OAuth)
• Styling: TailwindCSS 4 + shadcn/ui
• Language: TypeScript

Getting Started

Prerequisites

• Bun (or Node.js 18+)
• A Convex account
• A Firebase project with Google Auth enabled

Setup

1. Clone the repo:

   git clone https://github.com/Vochsel/dump.page.git
   cd dump
   

2. Install dependencies:

   bun install
   

3. Copy the environment template and fill in your values:

   cp .env.example .env.local
   

4. Start the Convex dev server:

   bun run deploy:convex:dev
   

5. Start the development server:

   bun run dev
   

6. Open http://localhost:3000

Environment Variables

See .env.example for the full list. Key variables:

Variable	Description
NEXT_PUBLIC_CONVEX_URL	Your Convex deployment URL
NEXT_PUBLIC_FIREBASE_*	Firebase project config values
FIREBASE_PROJECT_ID	Firebase project ID (used by Convex auth)
SCRAPINGBEE_API_KEY	Optional — for link metadata scraping

Scripts

Command	Description
bun run dev	Start dev server (uses .env.local)
bun run build	Production build
bun run lint	Run ESLint
bun run test	Run tests
bun run deploy:convex:dev	Start Convex dev server
bun run deploy:convex	Deploy Convex to production

MCP Server

Dump exposes an MCP server at https://www.dump.page/api/mcp so AI tools can read and write to your boards.

If you want the short version: Dump is a practical way to give ChatGPT, Claude, Claude Code, Codex, and similar tools the same project context instead of pasting links and notes into every conversation.

Installing Boards as Skills

Shared and public boards can also be installed with vercel-labs/skills via npx skills add.

Public board

npx skills add https://www.dump.page/b/<board-slug>

Shared board

npx skills add https://www.dump.page/s/<share-token>/b/<board-slug>

Dump serves a hosted SKILL.md for each board through the well-known skills endpoints. The installed skill is a snapshot of the board at install/update time. For live reads, search, and write access, use the Dump MCP server instead.

Quick Install

claude mcp add dump-mcp --transport http https://www.dump.page/api/mcp

Go to Settings → Integrations → Add More and paste:

https://www.dump.page/api/mcp

Go to Settings → Connectors → Add connector and paste:

https://www.dump.page/api/mcp

Go to Cursor Settings → MCP → Add new MCP Server. Use url type with:

https://www.dump.page/api/mcp

Or add to your .cursor/mcp.json:

{
  "mcpServers": {
    "dump-mcp": {
      "url": "https://www.dump.page/api/mcp"
    }
  }
}

Add to your VS Code settings (JSON):

{
  "mcp": {
    "servers": {
      "dump-mcp": {
        "type": "http",
        "url": "https://www.dump.page/api/mcp"
      }
    }
  }
}

Add to your ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "dump-mcp": {
      "serverUrl": "https://www.dump.page/api/mcp"
    }
  }
}

codex mcp add dump-mcp --transport http https://www.dump.page/api/mcp

Or add to ~/.codex/config.toml:

[mcp_servers.dump-mcp]
type = "http"
url = "https://www.dump.page/api/mcp"

Learn More

• Blog: how to share project context between ChatGPT and Claude
• Blog: the AI project handoff board
• Blog: MCP is better with shared context

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT