Give your agent the workspace it deserves.

Metagent turns scattered skills, MCP servers, and instructions into one tidy system—so your agents have room to succeed.

The Metagent app's overview: unused and never-used skill counts, reads per skill, token totals, and MCP connections

Agent setup should be inspectable, not archaeological.

Configured, installed, observed, and healthy are four different things. Metagent keeps them apart instead of flattening them into one status.

One core. Three ways in.

Hover to see what each one is for.

App

Browse every skill and server you have, see what's actually earning its place, and clear out the rest.

CLI

The same report as JSON, for any folder. Drop it in a script, a hook, or CI.

MCP

Your agent reads its own setup before it starts guessing. Read-only, always.

What lives where.

The app and CLI can change things. The MCP only ever reads.

CapabilityAppCLIMCP
Browse skills and servers
Full project report
Health findings
Real usage evidence
Utility scores ·
Duplicate detection · ·
Repair and remove ·

What your agent can see.

Point it at a folder and it reports on five things—metadata only, never the contents of your instructions.

Instructions

Which instruction files actually apply here.

Skills

What's in scope, and where it came from.

Servers

Which MCP servers this project is wired to.

Usage

What your agent has really been reaching for.

Health

What's worth fixing, without invented urgency.

Nothing leaves your Mac.

Metagent reads your agent logs and writes its analysis to a SQLite file on your own disk. No account, no server, nothing to upload.

Download