Blume
blume.codesEngineering
Not ranked yet—
Known for
Seeing what several coding agents are actually doing before one of them surprises you.
A desktop sidecar for coding agents, Claude Code and Codex and Cursor among them, showing what each one is doing, how much of your plan it has burned, and where its config files disagree with the instructions you gave it. Reads locally; history stays on the machine.
What it does7
- One view of every running coding agent and whether it has finished
- Tracks plan usage and token burn across Claude Code and Codex
- Surfaces the hidden config files that shape agent behaviour
- Flags mismatches between agent setup and your written instructions
- Proposes fixes for those mismatches (upcoming)
- Works across Cursor, Claude Code, Codex, omp and Pi
- Conversation history stays on the device
From Blume’s own site. Not used for ranking.
No data yet
Gets its first score next week.