SuperQode

The harness interoperability layer for coding agent

2 followers

The harness interoperability layer for coding agent

2 followers

There are many coding agent harnesses now, and meta-harnesses that route or switch between them by handing over the session. Model independence is nearly achieved. Harness independence is not, because harnesses do not talk to each other effectively. SuperQode is the interoperability layer that closes that gap, with agent to agent communication over A2A, ACP and UHP, and evaluation and optimization built for harnesses, so you know which one suits you or whether you need to build your own.

This is the 2nd launch from SuperQode. View more

SuperQode

Launching today

The harness interoperability layer for coding agents

Free

Launch tags:Developer Tools•Artificial Intelligence•GitHub

Launch Team / Built With

Viktor.comAn AI coworker that actually does the work

Promoted

Maker

📌

Hi Product Hunt. Shashi here. We launched SuperQode v1 here in January, and it was a different product. The plan was a software factory starting with Quality Engineering : agents covering the roles a software team actually has, including quality engineering and development engineering, working the whole delivery path instead of one step of it. The market moved underneath that plan a lot. Coding agents multiplied through the first half of the year, and the decision has changed being which model to use. It became which harness to run your workflow on. Every harness handles context, tools, permissions, memory and cost differently, and those differences decide whether a task completes or burns tokens and ships nothing. We already supported a wide range of models, so the harness was the new ground. We found the Meta-Harness work out of Stanford, built evaluation and optimization on top of it, and published our own metaharness package so the techniques run utside SuperQode too. By then two things were obvious that Anthropic and OpenAI had captured coding agents or development with their own harnesses. Everyone else was shipping harnesses of their own. Building one more would have added nothing, so we built the layer above them. Then Databricks released Omnigent, covering the same concepts with more engineers, a larger budget and far wider distribution. Our position holds on the part harness switching leaves out, which is measurement. GEPA, HarnessBench and SkillOpt produce numbers on your own repository before you commit to a harness. Not a vendor benchmark. Yours. Several companies have since launched cloud harness switching, which tells us the layer was the right place to stand. What v2 ships: - 104 harnesses in the Hub, 64 open source, switchable mid-session without starting over. - Full Agent2Agent. One published Agent Card advertises JSON-RPC and HTTP+JSON across A2A 1.0 and 0.3, so a single document registers on Gemini Enterprise, Microsoft Foundry and Bedrock AgentCore. 100% of MUST and 100% of MAY against the official TCK maintained under the open Foundation. - A hosted agent at a2a.superqode.dev you can call right now without installing anything. Ask it which harness fits your constraints and it answers from the Hub. - HarnessSpec, so your existing AGENTS.md, CLAUDE.md, Cursor rules and Copilot instructions become a spec you own, with a drift check that gates CI. Thanks to everyone who tried v1 in January. This technology has changing with insane pace and this release will help us to keep going through this, tell me what you think? Have we achieved harness independence for you or need something else?

Report

1d ago

Previous SuperQode Launches

SuperQodeAgent-Native Coding harness for agentic quality engineering

Launched on January 30th, 2026