See everything your AI agent can reach.
Aevrin reads how an AI coding agent is set up: its MCP servers, skills, plugins and hooks, and the files, keys, commands and network it can reach. Then it gives the setup a score out of 100.
- Reads files in this projectExpectedOK
- Runs any shell commandNo allow-list−10
- Can read ~/.aws/credentialsA cloud key, from the shell−15
- Calls notes-syncNot scanned, so ungraded−5
The problem Agent Posture solves.
A coding agent can read your code, run commands, use MCP tools, touch your keys and reach the internet. Almost nobody can say what one agent is really allowed to do. Aevrin turns that into a list you can read and a score you can track.
- 1
Scan
- 2
Work it out
- 3
Show what to fix
What goes in
- How the agent is set up
- Its permissions and allow-lists
- Its MCP servers, skills and plugins
- Its hooks
- Files and keys it can reach
- What network it may use
What comes back
- A score out of 100
- How risky the setup is
- Everything the agent can reach
- Keys it can see
- Risky permissions
- Fixes, best first
Why you can trust what it says.
A score you can explain
Every agent starts at 100. Each point it loses has a name, such as a key the shell can read, so you know which setting cost it.
Paths, not just settings
Aevrin joins what an agent can run with what it can read. You see the route from the agent to a secret, not a flat list of permissions.
Grades from real scans
An MCP server your agent can call keeps the grade from its own scan. A server nobody has scanned is shown as ungraded, never as safe.
The other five checks.
Put Agent Posture to work.
Start on the free plan. No card, and nothing renews on its own. Or add Aevrin MCP and use it from your agent.