VividKit Crystal clear AI coding
VI
Choose Your Setup Get Started with AgentKit Continue with ClaudeKit Migrate from ClaudeKit to AgentKit Clean cutover from ClaudeKit What Is AgentKit? How AgentKit Works AgentKit Kits, Skills, Agents & Hooks AgentKit Runtime Support AgentKit Desktop App AgentKit Configuration AgentKit CLI Quick Reference AgentKit Skills Cheatsheet AgentKit Workflows AgentKit Statusline Update AgentKit CLI & Kits AgentKit Helper Clean reinstall of AgentKit AgentKit Troubleshooting What is ClaudeKit? ClaudeKit CLI Guide Command Finder Commands Cheat Sheet Workflow Recipes CLI Quick Reference Claude Mechanics Permission Modes ClaudeKit Hooks Coexistence ClaudeKit x Codex CK → Other coding tools How ClaudeKit Works Inside ClaudeKit Finding Your Unknowns Session Recovery Fix from Logs UI/UX Guide Accounts Switcher IDE Config Remote Control Codex App System Diagramming UI Review Gate Promotions Support VividKit

Quick Ref / Command

Engineer Kit

/ak:use-mcp

Run existing MCP tools

Discover and execute existing MCP server tools through the narrowest runtime-native path or deterministic direct scripts, without copying credentials across runtimes.

01

Inventory native MCP

02

Choose narrowest path

03

Select smallest tool

04

Inspect schema

Rule 01

Prefer runtime-native MCP when the active session already exposes the server.

Rule 02

Use direct scripts for deterministic discovery and calls against .claude/.mcp.json servers.

Rule 03

Trust actual visible tool schemas over assumed Claude/Codex/Pi tool names.

Rule 04

Consume existing servers; use ak:mcp-builder only for creating new MCP servers.

Execution Map

Core lanes from input to output, aligned with the command's real execution path.

01

Start

Input, scope, route

  1. 1 Inventory native MCP Inspect MCP resources and tools already registered in the active Claude Code, Codex, or Pi runtime.
  2. 2 Choose narrowest path Use runtime-native MCP when visible; use direct scripts only for configured .claude/.mcp.json stdio servers or deterministic CLI workflows.
02

Work

Agent / skill execution

  1. 3 Select smallest tool Pick the smallest registered tool that directly satisfies the task and matches the user-requested mutation scope.
  2. 4 Inspect schema Validate required arguments and constraints from the live tool definition or the persisted assets/tools.json catalog.
03

Verify

Gate, review, validation

  1. 5 Classify effects Separate discovery, read-only access, external writes, destructive actions, credential use, uploads, and provider-cost calls.
  2. 6 Approve and call Confirm consequential mutations before execution, keep side effects inside scope, and report the server, tool, bounded arguments, result, and changed-state evidence.

Syntax, arguments, subcommands

How to invoke the skill: syntax, positional arguments, shared options, then each subcommand with its own syntax and outcome.

Syntax

/ak:use-mcp [task]

Arguments

[task] Required

MCP task

Natural-language request to discover MCP tools, inspect schemas, or call an existing server tool. Include the server name, read-only boundary, or approved external write when relevant.

/ak:use-mcp "List the issue tools available from the registered project server; do not call a mutating tool"

Special Operations

Runtime-native first

Lowest setup cost when the server is already registered in the active assistant runtime.

Direct scripts

Use cli.ts when you need explicit server names, JSON arguments, persisted schemas, or CI-friendly repetition.

Chrome profile work

Chrome DevTools MCP is profile-blind; invoke ak:chrome-profile first when cookies or an exact browser profile matter.

Sample Prompt

Concrete invocations for each flag, subcommand, mode, or route available in the live workflow.

Discover available tools Recommended 
/ak:use-mcp list the MCP tools available for the current runtime without calling mutating tools
Use when:
You need capability discovery before choosing a server/tool.
Expected:
Inventories the active runtime MCP surface first, keeps discovery read-only, then reports visible servers/tools or the visibility boundary that requires direct scripts.
Call a known tool 
/ak:use-mcp call the existing GitHub MCP tool to read issue 42 without changing the issue
Use when:
A server/tool is already registered and the task has a concrete read target.
Expected:
Selects the smallest matching GitHub tool, validates its live input schema, classifies the call as read-only, executes it, and reports the bounded arguments and result.
Use scripted fallback 
/ak:use-mcp discover tools from .claude/.mcp.json for this project and persist the catalog
Use when:
The runtime does not expose a configured stdio server but the project MCP config exists.
Expected:
Uses the bundled scripts client for the project configuration, runs list-tools/list-prompts/list-resources as needed, and persists assets/tools.json for schema review before any call.
Guard Chrome profile access 
/ak:use-mcp inspect the Chrome DevTools MCP options for my logged-in browser profile before using cookies
Use when:
An MCP-backed browser task depends on real cookies or an exact Chrome profile.
Expected:
Recognizes the DevTools MCP profile boundary, routes profile setup through ak:chrome-profile first, then uses only the returned visible MCP/browser capability for scoped work.

Handled Scope

  • Inventory native MCP
  • Choose narrowest path
  • Select smallest tool
  • Inspect schema
  • Classify effects
  • Approve and call

Related skills

Next