Project decisions, close at hand
Give your assistant answers it can come back to.
A menu-bar inbox for your AI assistant’s questions. Read the full context, make a choice, and keep the answer with the project.
Waiting in your menu bar.
Recent projects. Outstanding questions.
A direct route to your next answer.
Where should the first internal preview run?
The context
Field Notes is an example web app for six teammates. They need to try it this week. It has sample data and no sign-in yet, so the preview must stay on the company network.
You’re choosing a place for this first review. A permanent hosting setup can wait.
Answer saved
Illustrative project and question. Answers here stay in this page until you reload it.
From question to a decision you can find
A short conversation.
A lasting reference.
Your assistant asks.
Connect a project through MCP. Your assistant creates a named questionnaire, with the facts, constraints, and tradeoffs you need to decide.
You make the call.
Open a question from the menu bar. Choose an option, write your own answer, add notes, or explicitly leave it open. Drafts save as you work.
Both can return to it.
Review and submit to preserve the questions and answers together. Copy a reference, search past decisions, or let your assistant read the submission.
Fits around the work
One project.
One place for its decisions.
Each project folder has one session, with as many named questionnaires as you need. Recent projects and outstanding questions are a menu-bar click away.
Question definitions can change without erasing what you answered before. Submitted snapshots keep their original context.
Forgot a shortcut?
Hold Shift alone for two seconds in the app to see the key guide. Release it to return to your work.
Where are my questions and answers stored?
In JSON files inside your project’s .questionnaire folder. DecisionTray has no account or remote synchronization. Your connected assistant can read answers through MCP, so its own data policies still apply.
Does DecisionTray run an AI chat?
You keep using your existing assistant. DecisionTray provides a local stdio MCP server for creating and revisiting project questions, plus the desktop app where you answer them.
What if I’m not ready to decide?
Leave a question explicitly open. It stays visible in the inbox, and a submitted questionnaire records that the decision is still outstanding.
Can an assistant overwrite an answer?
Assistants can revise questions, but cannot submit human answers through MCP. Revisions preserve the previous draft or submission. Concurrent edits are checked, and a conflicting save keeps your unsaved input visible.
What does setup involve?
The current local build targets Apple Silicon Macs. Add the stdio MCP command to your assistant’s configuration with the project folder, then open that same folder in DecisionTray. The current MCP requires Node 26. Public installation instructions will accompany the download.
Release status
Public download coming soon.
DecisionTray is in local testing on macOS. A public installer and release date are not available yet.