Skip to content

feat: MCP server integration#124

Open
mcllerena wants to merge 20 commits into
mainfrom
ml/mcp-server
Open

feat: MCP server integration#124
mcllerena wants to merge 20 commits into
mainfrom
ml/mcp-server

Conversation

@mcllerena

@mcllerena mcllerena commented Apr 22, 2026

Copy link
Copy Markdown
Contributor

This is a first approach for the implementation, it'll be updated accordingly.

@codecov-commenter

codecov-commenter commented Apr 22, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.55%. Comparing base (3003e52) to head (a11b6a3).

Additional details and impacted files
@@           Coverage Diff           @@
##             main     #124   +/-   ##
=======================================
  Coverage   95.55%   95.55%           
=======================================
  Files           8        8           
  Lines        1823     1823           
=======================================
  Hits         1742     1742           
  Misses         81       81           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@mcllerena mcllerena requested a review from pesap April 22, 2026 15:55
@github-actions github-actions Bot removed the tests label Apr 23, 2026

@pesap pesap left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would like the MCP server to feel like something I can type from memory. Right now the proposed package name makes the common path clunky because users have to remember both the package name and the executable name: uvx --from mcp-server-plexosdb plexosdb-mcp.

Can we publish this as a separate adapter package named plexosdb-mcp, with the executable also named plexosdb-mcp? Then the happy path becomes uvx plexosdb-mcp, the safe path is uvx plexosdb-mcp --read-only, and users can run simple smoke checks such as uvx plexosdb-mcp health, uvx plexosdb-mcp --version, and uvx plexosdb-mcp --help without needing to know package/install plumbing.

I would still keep plexosdb and the MCP server as separate packages. plexosdb should remain the core library. plexosdb-mcp should be the thin adapter that depends on plexosdb and fastmcp. That keeps core installs clean while making the MCP install path obvious.

For local development, the docs can be explicit about the nested package path, for example uv run --project src/plexosdb-mcp plexosdb-mcp health and uv run --project src/plexosdb-mcp plexosdb-mcp --read-only. The current local uvx --from . plexosdb-mcp shape should not be the recommended path from the repo root because it installs the root package, not the nested MCP package.

Please update the package/docs around this user-facing shape before merge, and add a smoke check that verifies the documented command works.

@pesap

pesap commented Jun 10, 2026

Copy link
Copy Markdown
Collaborator

One more request from the agent-ergonomics angle: please make the CLI contract explicit. Data-bearing commands should support --json or --output json, diagnostics should go to stderr, failures should have stable exit codes and structured error fields where practical, and non-TTY use must not hang on prompts. A small metadata path such as plexosdb-mcp doctor --json, plexosdb-mcp version --json, or plexosdb-mcp capabilities --json would make it much easier for agents to discover and validate the installed server.

@mcllerena mcllerena requested a review from pesap June 17, 2026 23:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants