11 coding harnesses ship 380 tools, but only 124 are actually unique
yb2698 · x · 2026-10-07
To figure out how to improve coding harnesses, the author dissected the toolsets of 11 mainstream coding agents (opencode, pi, qwencode, kimi-code, codex, claude code, openhands, hermes agent, tau, deepseek harness, openclaw).
Key findings:
- The 11 harnesses integrate 380 tool entries in total, but only 124 are unique after functional dedup;
- Overlap was measured with a two-level "domain → job" taxonomy: e.g., the bash tool maps to a fine-grained job like shell.execution/run;
- All harnesses except codex overlap with the larger set by at least 65%, revealing massive redundant implementations across the ecosystem;
- The author concludes that most harnesses' capabilities (tools) could be centralized and shared instead of reimplemented.
More from coding & agent
- Agents love tidying up files nobody asked them to touch — JFPuget · 2026-10-08
- Open-source bridge turns a browser chat tab into an OpenAI-compatible API endpoint — harshanacz · 2026-10-08
- GraphRAG Bug: Deleted Documents Stay Indexed and Retrievable After Updates — JeremyCMorgan · 2026-10-08
- Paper: Vibe Coding Kills Open Source as AI-Recommended Repos Lose Stars — soumitrashukla9 · 2026-10-08
- Every publishes definitive guide to Compound Engineering, the philosophy behind 7k-star plugin — every · 2026-10-08
- Claude Haiku 5.5 matches GPT-6 Luna pricing but a stingier tokenizer hides a 1.25x cost hike — Simon Willison · 2026-10-08