Problem
semble install does not detect Command Code (cmd), requiring users to manually configure the MCP server. Command Code fully supports MCP servers via cmd mcp add and stores config in ~/.commandcode/mcp.json under the mcpServers key — the same stdio format used by most other supported agents.
Why semble
Command Code is a terminal-based AI coding agent with a growing user base. Auto-detection would let users run semble install and select Command Code alongside their other agents, consistent with how Claude Code, Codex, Cursor, Gemini CLI, Kiro, OpenCode, GitHub Copilot, VS Code, Windsurf, Zed, Reasonix, and Pi are handled today.
For Command Code specifically, the benefit is significant — it spins up an explore agent that reads the majority of files in a repo for each task, consuming thousands of tokens. Semble's semantic search would let it find relevant code in ~2% of the tokens instead.
Manual setup (current workaround)
Users can currently add semble manually:
cmd mcp add semble --scope user -- uvx --from "semble[mcp]" semble
Or with a direct binary:
cmd mcp add semble --scope user -- /path/to/semble
Proposed solution
Add Command Code as a detected agent in semble install, following the same pattern as existing agents (Claude Code, Cursor, Codex, etc.).
I'm happy to implement this.
Problem
semble installdoes not detect Command Code (cmd), requiring users to manually configure the MCP server. Command Code fully supports MCP servers viacmd mcp addand stores config in~/.commandcode/mcp.jsonunder themcpServerskey — the same stdio format used by most other supported agents.Why semble
Command Code is a terminal-based AI coding agent with a growing user base. Auto-detection would let users run
semble installand select Command Code alongside their other agents, consistent with how Claude Code, Codex, Cursor, Gemini CLI, Kiro, OpenCode, GitHub Copilot, VS Code, Windsurf, Zed, Reasonix, and Pi are handled today.For Command Code specifically, the benefit is significant — it spins up an explore agent that reads the majority of files in a repo for each task, consuming thousands of tokens. Semble's semantic search would let it find relevant code in ~2% of the tokens instead.
Manual setup (current workaround)
Users can currently add semble manually:
cmd mcp add semble --scope user -- uvx --from "semble[mcp]" semble
Or with a direct binary:
cmd mcp add semble --scope user -- /path/to/semble
Proposed solution
Add Command Code as a detected agent in
semble install, following the same pattern as existing agents (Claude Code, Cursor, Codex, etc.).I'm happy to implement this.