Hi Product Hunt — I’m Ryder, the maker of Meldwork.

Meldwork is a local-first desktop workspace for bringing the agent tools you already use into a shared task.

The question behind the product is simple: when several agents contribute to a decision, how do you keep the context, the evidence, and your own judgment together?

There are three ways to work:

Direct conversations for focused work with one agent.
Concurrent Responses for independent suggestions based on the same task snapshot.
Auto Discussion for proposals, cross-agent challenges, coordinated work, synthesis, and verification.

You choose the participants, provide the context, and review the result. The task keeps its contributions, supporting material, run state, and adoption decision together. Workspace writes are opt-in.

The demo uses a fictional launch plan so you can see the inputs and inspect the recommendations. Multiple agents can still miss things or agree on a weak assumption; review remains part of the work.

A few practical details: this is an Apple silicon macOS preview and requires compatible agent CLIs. Conversations and run records stay local, but the agents may send inputs to their configured model providers. The source is available for non-commercial use under the project license; commercial use needs prior written permission.

I’d love to hear where a second perspective is useful in your work. What would you want another agent to challenge before you adopt a result?