Don't lock into coding harnesses: build agent capabilities behind MCP interfaces
mostly_deterministic · reddit · 2026-09-28
A Reddit post argues that with Cursor, Claude Code, Codex, and opencode trading market share every few months, maintaining harness-specific cursor rules and plugins violates "program to an interface, not an implementation."
The author's proposal: treat frontier coding harnesses as the implementation and build your agentic capabilities behind a stable interface — make the harness swappable. Concretely: build local MCP servers or CLIs instead of harness-bound skill plugins; split the conversational harness from durable workflow/tool logic; use LangFuse for observability; a STDIO MCP server needs minimal boilerplate and distributes via your language's ecosystem. Harness-specific features stay as thin icing, while the complex capabilities you care about remain portable. The author reports this MCP-as-seam approach working well and sees early convergence in the ecosystem.
Related event: Developers Call for Decoupling Coding Skills from Harnesses via MCP(2 posts)→
More from coding & agent
- AI agent digs into game code, records its own footage to cut a promo trailer — AIandDesign · 2026-09-28
- JevControl open-sources a tool to test if smaller models can run your agent's routine choices — airesearch12 · 2026-09-28
- Claude Code CLI connects to Gmail easier than Google's own Antigravity, dev finds — randal_olson · 2026-09-28
- Pelican test on MiMo 2.6: plan mode cuts agent calls to 12-17 vs heavy iteration — PilgrimofHaqq2 · 2026-09-28
- Memory history is not reasoning history: agents lack a causal provenance layer — puppy_lover_2021 · 2026-09-28
- mark_k: AI coding is approaching a step change — systems only AI can debug — mark_k · 2026-09-28