Skip to content

fix(mcp): cli-core 0.707.0; mcp serve --profiles all with no profile serves its tools - #389

Merged
leemour merged 4 commits into
mainfrom
fix/parity-core-0707
Nov 19, 2026
Merged

leemour merged 4 commits into
mainfrom
fix/parity-core-0707

Conversation

@leemour

@leemour leemour commented Nov 19, 2026

Copy link
Copy Markdown
Member

Task

zoom-cli fails cli-core's release parity at 0.707.0 (run 44463906253) and sits on cli-core's known-broken.json. Make it pass, so it can come off the list.

Why

cli-core #3108 (0.698.0) changed mcp serve --profiles all on a computer with no profile: it now starts, answers initialize and tools/list, and refuses each tools/call with how to set up the first profile. Directories such as Glama start the server in an empty container to read its tools. The test says so when zm mcp serve has no profile to serve still expected exit 3, so the server waited on stdin and the test timed out after 15 s.

What

  • The test now connects an MCP client to zm mcp serve --profiles all with no profile (temporary HOME, memoryKeyring()), checks the tools all require profile, and checks a call is refused with "no zm profile on this computer yet".
  • cli-core 0.707.0 (was 0.691.0), cli-base 0.124.0 (was 0.109.0); contracts/wirecat.json regenerated.
  • Changelog fragment under Fixed. No release: the next regular one carries it.

Testing

Locally on this branch: pnpm typecheck, pnpm generate, pnpm docs:check, pnpm lint, pnpm lint:commands, pnpm test (91 files, 762 tests pass).

🤖 Generated with Claude Code

…serves its tools

cli-core 0.698.0 (#3108) made `mcp serve --profiles all` start on a computer with no profile:
it answers initialize and tools/list, and refuses each tools/call with how to set up the first
profile. The zoom test still expected exit 3, so it waited on stdin until it timed out in the
cli-core parity run. It now connects a client, lists the tools and checks the refusal.

cli-core 0.707.0 (was 0.691.0) and cli-base 0.124.0 (was 0.109.0); contracts regenerated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
leemour and others added 3 commits November 19, 2026 10:57
cli-core 0.698.0 and 0.705.0 added ydisk and amo to the first screen's "Works with" list, so
the help contract snapshot changed with the bump.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… 0.707.0

cli-core 0.705.0 moved the store to version 71, so the first list after an upgrade from 0.2.1
says schema 71, not 69.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The generator writes the Responses map that cli-core 0.706.0's api subgroups read, and
generate:check failed on the stale file.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@leemour
leemour merged commit cd3267a into main Nov 19, 2026
11 checks passed
@leemour
leemour deleted the fix/parity-core-0707 branch November 19, 2026 10:12
leemour added a commit to WireCatLabs/cli-core that referenced this pull request Nov 19, 2026
…scheduled cancel|edit|send (#3146)

## Task

`scripts/parity/known-broken.json` should list only CLIs that are really
broken, each with a reason. The 0.707.0 release run
([44463906253](https://github.com/WireCatLabs/cli-core/actions/runs/44463906253))
shows most of the list passes.

## What changes

| CLI | Before | Now | Evidence |
| --- | --- | --- | --- |
| mail-cli, cli-memo, tg-cli, gcal-cli (both entries), bitrix24-cli |
listed (#2896, #2903) | removed | the run's "passes; remove it" warnings
|
| cli-template | listed (#2896) | removed | its tests pass in the run's
`template` job |
| max-cli | listed (#2896) | **stays until 2026-11-21**, re-pointed to
this PR | fixed in WireCatLabs/max-cli#984, but max's own CI runs the
parity manifest check against cli-core on npm, so #984 merges only after
a release carries the rows below. Until then a release builds max's
main, which still fails |
| drive-cli | listed (#2896) | removed | fixed in
WireCatLabs/drive-cli#384 |
| zoom-cli | listed (#2896) | removed | fixed in
WireCatLabs/zoom-cli#389 |
| whatsapp-cli | listed (#2896) | **stays**, re-pointed to #3139 with a
reason | its `messenger.test.ts` fails: #3139 added `cancelScheduled`,
`editScheduled`, `sendScheduled` to `COMMAND_NEEDS`, and
`WHATSAPP.lacks` does not name them yet |

An entry now has a `reason` field; the scripts read only `repo` and
`until`, so nothing else changes. The error message asks for the reason
too.

## A gap this uncovered: `parity.json` rows

#3139 added `messages scheduled cancel|edit|send` without `parity.json`
rows. The manifest check (`main (max)`, `main (tg)`) failed in the
release run, but `continue-on-error` hid it because both CLIs were
listed. Removing them without the rows would make the release gate fail.
The rows say: in max and tg, not in wa ("WhatsApp has no scheduled
messages"); `--html` on edit in tg only, as on `messages edit`.
`docs/dev/STANDARD.md` re-rendered.

## Testing

- The parity check against the release run's own `commands.json`
artifacts: `tg`, `max`, `wa` and `wording` pass with the new rows; `tg`
and `max` failed before them.
- `pnpm vitest run src/parity` (53 tests), `pnpm docs:check`.
- This branch is named like the CLI pull requests, so this PR's parity
builds max-cli, drive-cli and zoom-cli from their fix branches
(`app-ref.sh`). A PR's parity runs no CLI tests; those ran locally in
each CLI and run at the next release.

Merge freely: max-cli stays listed, and zoom-cli and drive-cli merge on
their own green CI. The release after this one should run with
`--app-ref fix/parity-core-0707` so parity builds max-cli from its fix
branch.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant