Aller au contenu principal
Integration guide

MCP or CLI — same power, different use cases

MCP and CLI expose the same 173 GAM operations. What changes: MCP loads a full AI context (~30K tokens), CLI runs a targeted command (~40 tokens). The right tool depends on your situation.

Two tools, one objective

MCP AI interface

Loads 173 tools into AI context. The agent reasons, orchestrates, and guides you — ideal for anything requiring judgment or multiple nested steps.

Use when

  • You're exploring your inventory or discovering GAM
  • A campaign requires multiple nested steps
  • You want a preview before every write action
  • You're using Claude, Codex, Gemini or ChatGPT
CLI Lightweight layer

Runs a targeted operation, returns JSON, consumes zero context tokens. Ideal for scripts, internal Skills, and quick lookups.

Use when

  • You need a quick result without spinning up an AI agent
  • A skill needs to fetch data mid-MCP workflow
  • A cron job or CI/CD pipeline runs unattended
  • Context is limited and 173 tools are too heavy

The real difference: context cost

~30,000 tokens
at MCP startup
173 tool descriptions loaded into the AI prompt
~40 tokens
per CLI call
Just the command + the JSON result

My situation, my tool

Built to work together

MCP is the primary interface. CLI is the lightweight layer Skills use when context matters. Same account, same credits, same guardrails.

MCP
Full AI control
CLI
Zero-token lookups
Optimized context
Same account · same credits

What about Skills?

MCP and CLI are transports — they provide raw capabilities. Skills add operational context: step order, checks, business rules. Without Skills, the AI improvises. With Skills, it's immediately operational.

Discover the 9 Skills
Set up MCP Install the CLI See integration methods