Skip to main content

Tools Management

Control which MCP tools are available to AI agents, organized by risk level.

How It Works

MCP Exposes Tools
get_repos → read
create_issue → write
delete_repo → destructive
QuilrAI Categorizes
Low risk: 5 tools
Medium risk: 3 tools
High risk: 1 tool
Admin Controls
get_repos: enabled
create_issue: enabled
delete_repo: disabled
QuilrAI
  1. MCP Exposes Tools - The server declares its available tools
  2. Gateway Categorizes - Tools are sorted by risk level automatically
  3. 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.

ToolDescriptionStatus
search_documentsSearch the knowledge baseEnabled
delete_workspacePermanently delete a workspaceDisabled

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.