Tools Management
Control which MCP tools are available to AI agents, organized by risk level.
How It Works
- MCP Exposes Tools - The server declares its available tools
- Gateway Categorizes - Tools are sorted by risk level automatically
- Admin Controls - Enable or disable each tool individually
Tool Categories
Read Only - Low Risk
Tools that only read data. Safe to enable by default - no modifications to external systems.
Examples: get_file, list_repos, search_docs, read_database
Write Access - Medium Risk
Tools that create or modify data. Review before enabling - changes can be undone but may have side effects.
Examples: create_issue, update_record, send_message, write_file
Destructive - High Risk
Tools that delete or irreversibly modify data. Disabled by default - enable only when explicitly needed.
Examples: delete_repo, drop_table, revoke_access, purge_data
Per-Tool Controls
Each tool shows its name, description, and an enable/disable toggle. Click "View schema" to inspect the tool's JSON input schema.
| Tool | Description | Status |
|---|---|---|
search_documents | Search the knowledge base | Enabled |
delete_workspace | Permanently delete a workspace | Disabled |
Seamless and Transparent
Disabled tools are hidden from AI agents automatically. When an agent calls tools/list, only enabled tools are returned. No code changes required on the agent side.