Simulation run on this device. Scenario values are not sent anywhere — the engine is fully local.
agents studio · pro
MCP Inspector Lite
Walk a mock server's capabilities, schemas, and permission surface.
Which tools does this MCP server really expose?
Engine 1.0.0 · studio
- Tools
- 5
- Broad scope
- 3
- Ambiguous
- 1
| Tool | Description | Verdict |
|---|---|---|
| read_file | Read any file on the host | broad scope |
| write_db | Insert, update, or delete any row | broad scope |
| x | Does stuff | ambiguous description |
| admin_exec | Run shell commands as root | broad scope |
| search | Search the knowledge base | ok |
- highBroad scope: read_file
Write/delete/admin access should be split into narrow, auditable tools with explicit inputs.
- highBroad scope: write_db
Write/delete/admin access should be split into narrow, auditable tools with explicit inputs.
- mediumAmbiguous tool: x
Tool-choice models confuse this with neighbours — write a one-line contract that names inputs and effect.
- highBroad scope: admin_exec
Write/delete/admin access should be split into narrow, auditable tools with explicit inputs.
Split broad tools, rename vague ones — tool-choice accuracy is mostly description quality.
Method
- Audits each tool's name and description for broad scope (write/delete/admin/exec), ambiguous descriptions, and sensitive surfaces.
- Tool-choice accuracy is mostly description quality, so ambiguous tools are flagged even when their scope is narrow.