fix(mcp): give 'mcp add --command' a distinct argparse dest

[discodirector]

What does this PR do?

hermes mcp add <name> --url <url> silently launches an interactive chat session instead of registering an MCP server, because of an argparse dest collision. This PR fixes the collision and adds a regression test at the parser layer.

The top-level subparsers in hermes_cli/_parser.py use dest="command" so the dispatcher in main.py can route on args.command. The mcp add subparser registers a --command flag (the stdio command for an MCP server) without an explicit dest=, so argparse derives dest="command" from the flag name. Whenever --command is omitted (e.g. on the documented --url install path), the subparser still writes args.command = None, clobbering the top-level "mcp". The dispatcher then sees args.command is None and falls through to cmd_chat. cmd_mcp_add is never reached — there is no error, no warning, no log.

The fix gives the flag an explicit, non-colliding dest. The user-facing CLI flag --command is unchanged; only the in-memory namespace attribute moves.

Related Issue

Fixes #19785.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• hermes_cli/main.py — declare mcp_add_p.add_argument("--command", dest="mcp_command", ...) so the flag no longer overwrites args.command. Comment explains the constraint so a future cleanup pass does not re-introduce the collision.
• hermes_cli/mcp_config.py — cmd_mcp_add reads getattr(args, "mcp_command", None) instead of args.command. Comment cross-references the parser.
• tests/hermes_cli/test_mcp_config.py — _make_args helper now defaults mcp_command=None (was command=None); the four call-sites that asserted stdio-server behavior pass mcp_command="npx" / mcp_command="uvx" accordingly. Pure mechanical rename, no logic change.
• tests/hermes_cli/test_mcp_add_command_dest.py (new) — three parser-level regression tests modelled on test_argparse_flag_propagation.py and test_subparser_routing_fallback.py. They build a minimal replica of the parent + mcp add subparser, parse representative argv vectors, and assert (a) args.command stays "mcp", (b) --command npx populates args.mcp_command (not args.command), (c) the bare-mcp add form also preserves the top-level dest.

How to Test

1. Reproduce on main at cfd86dc:

   hermes mcp add example --url 'https://example.com/mcp'
   

   Observe an interactive chat prompt opens; hermes mcp list shows no example entry.

2. Apply this PR. Repeat the command. The MCP server is now registered; hermes mcp list shows example.

3. Run the new regression test:

   pytest tests/hermes_cli/test_mcp_add_command_dest.py -v
   

   All three tests should pass. Reverting the dest="mcp_command" change makes test_url_invocation_preserves_top_level_command and test_bare_mcp_add_does_not_clobber_command fail with args.command == None.

4. The existing tests/hermes_cli/test_mcp_config.py suite (covering stdio-add, env vars, presets, etc.) continues to pass after the helper rename.

Notes on scope

• No fallback read on the old args.command attribute. A naïve fallback (getattr(args, "mcp_command", None) or getattr(args, "command", None)) would silently return "mcp" whenever cmd_mcp_add is invoked through the real dispatcher (because argparse sets args.command = "mcp" from the top-level subparser), making cmd_mcp_add think the user passed --command mcp. The clean rename is safer.
• The --command CLI flag name is preserved; this is purely an internal namespace change. No config-file format changes, no behaviour change for any path other than the previously broken one.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits
• I searched for existing PRs — no duplicate
• My PR contains only changes related to this fix
• I've run pytest tests/ -q and all tests pass — I ran tests/hermes_cli/test_mcp_add_command_dest.py in isolation (3/3 pass) and verified Python syntax of the four touched files. I was not able to run the full pytest suite in my dev environment; happy to address any CI failures.
• I've added tests for my changes
• I've tested on my platform: Windows 11, Python 3.12

Documentation & Housekeeping

• I've updated relevant documentation — N/A (no user-facing docs reference the internal dest name; the --command CLI flag is unchanged)
• I've updated cli-config.yaml.example if I added/changed config keys — N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — N/A
• I've considered cross-platform impact — argparse behaviour is platform-independent
• I've updated tool descriptions/schemas if I changed tool behavior — N/A

[alt-glitch]

Note: #17294 is an earlier PR for the same argparse dest collision fix. This PR appears more complete (includes regression tests).

[alt-glitch]

Note: #17294 is an earlier PR for the same argparse dest collision fix. This PR appears more complete (includes regression tests).

[teknium1]

Merged via #21204 — your commit was cherry-picked onto current main with your authorship preserved (commit 4f364c4). Thanks for the clean root-cause analysis and the parser-level regression test — made the salvage trivial. Sorry it bit the Sonoglyph onboarding; the fix is on main now and next hermes update will pull it in. https://sonoglyph.xyz is dope, by the way.

[teknium1]

Already merged via #21204 (commit 4f364c4 on main) — your commit landed with your authorship preserved via rebase-merge. E2E-verified just now that hermes mcp add <name> --url ... now dispatches to cmd_mcp_add correctly, writes the server to config.yaml, and hermes mcp list shows it. Thanks for the clean diagnosis, minimal repro, and regression test.

[discodirector]

Merged via #21204 — your commit was cherry-picked onto current main with your authorship preserved (commit 4f364c4). Thanks for the clean root-cause analysis and the parser-level regression test — made the salvage trivial. Sorry it bit the Sonoglyph onboarding; the fix is on main now and next hermes update will pull it in. https://sonoglyph.xyz is dope, by the way.

Thanks ser for the trust, now onboarding really should go smooth as butter.
Waiting for your own Sonoglyph artifact!