Overview

The short version, computed from the current data.

Harness coverage by tool

How many of the tracked capabilities each tool supports. Dark segments are native or compatible support. Hatched cells haven’t been verified yet.

    Coverage by area

    Share of each area's capabilities a tool supports natively or compatibly. Select a cell to compare that area.

    Compare

    Pick the tools your team uses to see what you can write once, what needs a per-tool adapter, and what only one of them offers. The selection is kept in the URL, so you can share it.

    Tools

    What stands out about each tool, and what will catch you out.

    What changed

    Every review is recorded. Changes to support levels are listed below. Edits to notes and sources are counted but not listed. Wording-only edits aren’t counted.

      Sources & data

      How the data is gathered, what the levels mean, and how to reuse it.

      Levels

      Priority tiers

      P0, P1
      The portable core: a shared standard or something every tool does. Write it once.
      P2
      Same idea everywhere, different files or fields. Generate it per tool with a thin adapter.
      P3
      Nice to have, or specific to one tool.

      How it’s researched

      Each tool has a long-form profile in my research notes, checked against the vendor’s documentation and, where the docs are vague, against the tool itself. Every cell cites a vendor URL or the section of those notes it came from, such as tools/cursor.md §7. In the notes, “KB” means my research notes and “hub” means their summary matrix.

      If something is undocumented or the sources disagree, it stays unverified. I don’t guess. A tool that runs in several places, such as a CLI, an IDE and the cloud, is marked partial when only some of them have a feature, and the note names which.

      The page is regenerated from the notes after each review, so the date at the top is the date of the last review, not the last deploy.

      Download

      • Full data (JSON)Tools, capabilities, every cell with notes and sources, and the changelog.
      • One row per tool and capability.
      • Plain text (Markdown)Readable tables, also the version served to AI agents.

      Cite