Install once, then describe the task
The public dcc-mcp Skill contains the operating procedure. dcc-mcp-cli handles Gateway status, tool discovery, typed calls, diagnostics, updates, and Marketplace installation.
Run this from the workspace used by Codex, Claude Code, Gemini CLI, GitHub Copilot, Cursor, Windsurf, OpenCode, Cline, Roo Code, Kiro CLI, Amp, or another Agent Skills-compatible host:
npx --yes skills@1.5.23 add dcc-mcp/dcc-mcp-agent-plugins --skill dcc-mcpUse --global for a user-level installation. Start a new agent session if the host discovers Skills only at startup. Native plugin and registry installation options remain available in the dcc-mcp-agent-plugins repository.
Keep the Skill and CLI current
Use the updater that owns each installation. Agent Skills installed from GitHub use the skills lock file:
# Update managed copies in this workspace.
npx --yes skills@1.5.23 update -p -y
# Update managed user-level copies.
npx --yes skills@1.5.23 update -g -yCheck the DCC-MCP CLI separately, then apply the verified update only with consent:
dcc-mcp-cli update check
dcc-mcp-cli update applyOpenClaw and direct ClawHub installations use their own recorded source and update commands. See Keep installed Skills current. Do not add --force to unattended updates; pinned or locally modified Skills should stop for review.
Use a short prompt
The Skill carries the procedure, so the prompt only needs the task and its safety boundary:
Use the dcc-mcp Skill to <describe the DCC task>. Ask before installing or changing system state, and finish with verification evidence.Install only the Skill you need
| Intent | Skill |
|---|---|
| Operate a live DCC, discover tools, or search extensions | dcc-mcp |
| Create or modernize a complete adapter/runtime | dcc-mcp-creator |
| Create or improve a DCC-specific Skill package | dcc-mcp-skills-creator |
Replace the --skill value in the universal command only when the task matches one of the two developer routes.
Search, describe, and call
# Confirm the gateway and connected hosts.
dcc-mcp-cli health
dcc-mcp-cli list
# Search narrowly, then follow the returned next_step.
dcc-mcp-cli search --query "create sphere" --dcc-type maya
# Execute the returned tool slug with validated arguments.
dcc-mcp-cli call <tool-slug> --json '{"radius": 2.0}'Do not guess tool names or load every backend schema. Search first, then perform the load or describe step named in next_step.
Use the request ID before retrying
Keep the request_id from a failed call. Use the CLI's built-in evidence path:
dcc-mcp-cli doctor
dcc-mcp-cli stats --status failureThen discover dcc_feedback__report through the same search workflow. Review and redact the report before sharing it. Creating an external issue still requires user authorization.