The useful bits
- A single-harness project can qualify. Every published entry needs a documented integration and a concrete coding-workflow benefit.
- Filter against verified harness compatibility, purpose, and entry type. Compare requirements and limitations instead of scores or popularity.
- The original research remains intact. The curation ledger accounts for included, excluded, and unverified entries.
Coding assistants are powerful, but the surrounding workflow still matters: how a task is planned, what gets remembered, when tests run, and who reviews the result. Frameworks and add-ons tackle different pieces of that work. They should be compared on those pieces.
Browse the Atlas
Search by name or purpose, choose the coding harness you use, and narrow the directory by the work you need help with. Each result explains its purpose, verified compatibility, and installation format. Compare two to five entries side by side.
The live directory derives its counts from the published entries. Unsupported or unresolved claims stay out of those results.
What belongs in the directory
Both complete workflows and focused extensions qualify: reusable skills, rules, hooks, memory and context tools, review helpers, and orchestration layers. A project supporting just one harness is welcome when its benefit and installation are documented.
Standalone agents, general SDKs, infrastructure platforms, unrelated applications, generic connectors, directories, and duplicates are outside this directory’s scope.
Evidence before classification
The original research is preserved. Its old categories are starting points for review, not automatic inclusion decisions. The curation ledger records every original entry, including those still awaiting evidence.
Verification means checking upstream documentation and recording the source and date. It does not imply that every integration has been executed or independently benchmarked. Detail pages make those limits explicit.