Skip to content
| Marketplace
Sign in
Visual Studio Code>Linters>MLflow CompanionNew to Visual Studio Code? Get it now.
MLflow Companion

MLflow Companion

Gap Hunter Labs

| (0) | Free
Browses local MLflow tracking-store runs (params, metrics, tags) in a sidebar tree, reading mlruns/ directly from disk. No data leaves your editor.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

MLflow Companion (VS Code)

Browses local MLflow tracking-store runs (params, metrics, tags) in a sidebar tree, reading mlruns/ directly from disk — no tracking server, no MLflow UI, no network call.

v0.1, new niche. Not a port from the Gap Hunter Labs IntelliJ- family catalog. Evidence: MLflow's own web UI requires running mlflow ui (a local server) or mlflow server — there's no lightweight, always-there editor view of mlruns/ for a data scientist who just wants to glance at recent runs without leaving VS Code or starting another process.

Real format, not a guess

The mlruns/ directory layout and file formats this extension reads were verified against real, captured example directories found on GitHub (not a paraphrased description) — a run's meta.yaml, an experiment's meta.yaml, a params/<name> file, and a real 3-point metrics/<name> file — plus MLflow's own service.proto source for the numeric run-status codes (RUNNING=1 … KILLED=5). See the comment at the top of src/mlrunsParser.ts for exact citations.

What it does

An MLflow Runs view in the Explorer sidebar, populated by reading a top-level mlruns/ directory in your workspace (MLflow's default local "file store" backend — the layout you get without configuring a remote tracking server). Each experiment expands to its runs; each run shows its status (RUNNING/FINISHED/FAILED/etc.) and expands to its logged params, metrics (latest value + point count), and tags.

v0.1 scope, honestly noted

  • Local file store only. MLflow also supports a SQLite/Postgres/ MySQL-backed tracking store and a remote tracking server — this extension reads only the plain-directory mlruns/ layout used when no tracking URI is configured; neither the database backends nor a remote server's HTTP API are read.
  • Browsing, not comparison. Lists what's there; doesn't yet chart metrics over steps or diff params across runs.
  • Read-only. Never writes to mlruns/.

Development

npm install
npm run compile   # or: npm run watch
npm test

To build an installable package without publishing:

npx @vscode/vsce package

License

Apache License 2.0 — see LICENSE.

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
  • Your Privacy Choices
  • Consumer Health Privacy
© 2026 Microsoft