Brighter Roadmap for 2026

[iancooper]

Introduction

As its new year, I thought it might be worth talking about our roadmap for 2026. Some of this is under construction, some no one has looked at yet. The earliest we would release a V11 would be December 2026. But if we decided that we needed breaking changes, work might begin on that as early as May or June

V10

• Agent Friendly: Various steps lead us towards making it easier to work with agents. Our initial goal is really to make it easy to create transports and boxes for new middleware or db via an agent. @lillo42 is working on a standard test suite for these, and a generator to let you create them. I have working on what I now know should be Claude Skills for implementing transports and boxes. And we have been joined by @darrenschwarz. Hoping that we can make this really easy soon, and thus rapidly expand our support. For example we might want a skill package for setting up a new project with generated tests, etc.
• Aspire: We have had both @holytshirt and n ow @lillo42 look at this. Our goal is really to work better with the Aspire hosts, as we don't really need the clients, we are the clients. Some work on how we handle Configuration may help here, so that we can more easily work with Aspire
• Durable Execution: @iancooper (me) has been working on our durable execution support and some elements of it crept into V10's interfaces. This should enable you to run long-running workflows with Brighter. It is an alternative to approaches like Mass Transit's Sagas. We began this before agents made it fashionable, but we'll see if it helps agents once we have it
• Universal DLQ: DLQ support is not there for streams, and many folks would like it. We will add something shortly. You can use it for an error queue, although we don't feel that is best practice.
• Skills: I am separating this from Agent Friendly, although it is similar, because this is about an agent being able to write Brighter code. I had thought this was about an MCP Server, but I suspect that skill packages, along with tools are better. For example, if we had a dotnet new template, along with a Skills package, we might be quickly enable an agent to create a Brighter consumer.

Otherwise the backlog is there in the issues, all the items have value and we would love to support people picking them up.

We also need to pick up Darker V5, and its possible we will experiment using agents to speed this up: BrighterCommand/Darker#273

V11

• Configuration: This remains our biggest pain point for newbies and a new version would give us the ability to make more changes to how this is done. As much as possible we want to make it easy to get started, and accept defaults, but still give you plenty of power under the hood.
• Multiple Outboxes: This one comes up, that we can't support multiple outboxes. It gets us into the issues with our using statics, in tension with the simplicity of using statics. [Feature] Does Outbox supports multiple connections? #2104

Other than that, I am not sure what my force V11 and both of these could be resolvable without breaking anything.

Thoughts?