Skip to content
Vardan Verma
Let’s talk
All work01 / 08 · Developer tooling

AI Manager

A dashboard and MCP server for the sessions, questions and decisions in an AI-assisted development workflow.

Screen recording of AI Manager running on its built-in sample data: six coding agents at work, a question arriving under Needs you and answered from the dashboard so the session carries on, then the team under that session, usage left on each subscription, the Costs page views, and four colour themes.
Film · 0:27
Fact sheet
Tracks
Sessions, subagents, worktrees
Surfaces
Every question waiting on you
Reports
Usage and cost

01The problem

When several coding sessions are running at once, the work becomes difficult to follow. A question can sit unanswered in one session while another finishes or creates a new worktree.

02The engineering

I built AI Manager, a dashboard and MCP server that brings sessions, subagents and worktrees into one place. Usage and cost reporting provide context for the work.

Sessions
Subagents
Worktrees
MCP server
Dashboard
Question waiting
Answer
Session continues
Interactive model · sample data
Session control● Connected
API migrationWorking through the implementation
Running
Interface reviewA decision is waiting on you
Needs you
Test suiteChanges verified
Complete
Question from Interface review

Which theme should be the default?

03The outcome

Questions waiting for input are surfaced in the dashboard, with a path to answer and let the session continue.

Next3D productionpipeline
Have a similar problem?