Project Intelligence System keeps every project structured, logged, and fully readable by any AI tool. Switch between tools without losing a single step. Every AI that touches a project leaves context for the next one.
Switch AI tools at any point. The next tool knows exactly what was done, what is pending, and what the goal is.
Each AI that works on a project records what it did. That record becomes permanent context for every AI that comes after.
Dash. Claude. Cursor. Codex. Hermes. Connect any of them over MCP and it understands your project immediately.
Every project lives in a shared database. Switch AI tools mid-task and nothing is lost. The next AI picks up with full knowledge of what was done, what is next, and why.
Every time an AI works on a project it logs what it did, what problems it hit, and what it solved. That log becomes permanent context for every AI that comes after.
Start in Dash. Continue in Claude Code. Switch to Cursor. Each one connects to the same dashboard over MCP and instantly understands the full project state.
Describe any idea in plain language. The built-in agent plans every step, sets priorities, and creates a fully structured project record ready for any AI to work on.
Find an interesting tutorial, video, or article while browsing? Save the link. Come back later and ask any AI to explain it, break it down, or turn it into a project.
Steps, expected results, priorities, and agent history. Every project is fully documented and queryable so any AI that connects knows exactly where things stand.
Tell Dash to research your project and it searches the web, picks the best resources, and saves them directly to the project as reference links. Ready before you take your first step.
Mark projects as Urgent, Scheduled, or Someday. Set custom priorities with a single click. Move projects from active to paused to done to archived as they progress through their lifecycle.
Mid-conversation with an AI and something clicks? Tell it to save the insight. Notes attach directly to the project so the next AI, whether Dash, Claude, or any other, reads them as part of the full context.
Dash is a GPT-4o powered agent embedded in the dashboard. Describe any idea and it instantly creates a fully structured project with every step planned, priorities set, and context written so other AI tools can continue the work immediately.
After Dash bootstraps the project, open Claude Code, Cursor, or any MCP-connected tool. They connect to the same dashboard and already know the full plan. No copy-pasting. No re-explaining. Just continue.
This dashboard is an MCP server. Point Claude Code, Claude Desktop, Codex, Copilot, Cursor or a headless agent at it and you stop clicking: you say what happened and it updates the project. It reads your steps, session logs, expected results and priorities directly, so the context never disappears because it was never just in a chat window.
Connecting takes one message to your assistant. It reads how the dashboard works by itself the first time, so there is nothing to configure and nothing to explain.
Every step, log, and expected result lives in Supabase. Any AI that queries it has the same context as any other. The database is not storage. It is the shared brain.
Chat windows close. IDE sessions end. But the project record in the database never disappears. Open a new tool and the full history is already there.
When an AI works on a project it logs what it did. Those logs stack over time and become a rich, permanent context that makes every future AI session more effective.
Both live on the Agent Access page inside the dashboard. Pick the one that matches the agent you have.
No key is ever created, so there is no secret to lose. Cut it off from the same page whenever you want.
A machine with no browser can never finish a sign-in, so it gets a credential it can hold. One key per agent means revoking one leaves the rest alone.
A connected agent only ever sees the account it was connected to. Which account a request belongs to is decided by the credential it presents, never by anything the agent sends, so it cannot reach another person's projects even if it asks. Permissions are per connection, deleting anything needs an explicit confirmation, and every call an agent makes is listed in your dashboard, including the ones that were refused.
Dash skills are custom instruction sets stored in the database. Create a skill, give it a /command, and Dash reads it and applies it every time you invoke it. Any AI connected over MCP can read them too, so it plans the way you want without being told.
Each skill has its own /command. Type /human, /pm, /analyze — whatever you created. Dash reads the skill from the database and applies it to your message.
Connect Claude Code or Cursor over MCP and they read your skills before planning anything. The way you want work broken down travels with the project.
Create skills from the dashboard, modify the command, edit the instructions, toggle them on or off. Full control without needing a developer.
More production-grade applications across cybersecurity, AI, and developer tooling. Each one built to solve a real problem with real technology.
A full-stack application with a custom admin dashboard, role-based access control via Clerk, PostgreSQL triggers, an OpenAI-powered contact assistant, and Arcjet WAF protection. Built like a real product, not a static page.
An experimental honeypot that turns defense into a game. The Sentinel AI responds to attack attempts with evolving behavior. Heuristic sensors, a real-time War Room visualization, and a Wall of Infamy for persistent researchers.
This project is open source. The repository includes full setup instructions, environment variables, and database schema so you can run your own instance.
VIEW ON GITHUB