Agent Manager
A control panel for managing Claude agents from Obsidian.
A personal, in-progress tool — kept closed by choice, not unfinished portfolio filler.
The problem
As the number of custom AI agents grows, there is no single place to configure, inspect, and drive them — and no way to test the logic without booting a full editor around it.
The solution
An Obsidian plugin with an 11-page interface and a test-driven, editor-independent core (128 checks) so the logic runs and verifies without Obsidian. Live subscription sessions connect real agents end to end.
What it does
- ▹11-page UI for configuring and driving Claude agents
- ▹Editor-independent core with 128 automated checks
- ▹Claude-CLI subscription provider as the default backend
- ▹Live subscription sessions verified end-to-end
- ▹Private plugin, developed test-first
How it works
- 1
Separate the core
Agent logic lives in an Obsidian-free core so it can be unit-tested with Vitest — no editor required to prove correctness.
- 2
Drive from the UI
The plugin's 11 pages configure agents, models and providers, then dispatch real sessions.
- 3
Run live
Sessions connect through the Claude-CLI subscription provider and stream results back into the vault.
Architecture
Click a node to see what it does.
How it looks
Tech stack
- TypeScript
- Obsidian API
- Vitest
- Node
- esbuild