About this skill

Problem

When an agent needs to operate DCC instances from a shell environment, raw MCP JSON-RPC calls or hand-built curl requests can become brittle around protocol details, authentication, and tool parameters. dcc-cli-gateway turns the DCC-MCP REST gateway into a stable command-line control surface, letting the agent use dcc-mcp-cli to check instances, discover tools, inspect schemas, and invoke tools. It is especially useful when DCC host operations need to stay inside an auditable agent workflow instead of inventing one-off API calls.

How It Works

The skill prefers an existing dcc-mcp-cli binary on PATH. If it is missing, it asks for user permission before downloading the matching platform release; if the download fails, it falls back to the bundled Python REST helper. The helper reads the gateway URL from DCC_MCP_BASE_URL and prefers JSON for agent-facing interactions. The workflow is intentionally fixed:
- Run health and list first, then inspect total, dcc_type, stale rows, and the target instance_id
- Run search only when a non-stale target exists, and copy the returned slug exactly
- Use describe to read inputSchema and safety annotations before any call
- Pass tool-specific fields such as code, file_path, and radius inside the --json object rather than as top-level CLI flags

Limits

This skill is meant for CLI-based DCC-MCP agent access. It is not a substitute for direct Maya, Blender, or Houdini scripting, and it avoids raw MCP tool interfaces except when debugging the gateway itself. If list returns total == 0, the gateway is unreachable, or the user has not approved setup, the agent should stop and explain instead of installing packages, editing configs, or launching apps.

Screenshots

Use Cases

Before starting a DCC task, check gateway health, instance count, stale rows, and target instance ID.
When list shows a non-stale instance, search for tools, copy the slug, and describe its schema.
When invoking a DCC tool, put code, file_path, and radius inside the --json object.
If dcc-mcp-cli download fails, use the Python REST fallback to access the same gateway endpoints.

Best For

Automation engineers who need an agent to call the DCC-MCP gateway from shell.
Platform engineers integrating DCC hosts into agent toolchains with clear permission boundaries.
AI application developers who want auditable DCC tool calls instead of hand-written curl.
DevOps engineers maintaining DCC CLI workflows and CLI-download fallback handling.