DEV Community

Cover image for Assembly Is Evolving: A New Interface, Better Experiences, and Room for More Contributors
Anubhab Das
Anubhab Das

Posted on Edited on

Assembly Is Evolving: A New Interface, Better Experiences, and Room for More Contributors

Let intelligence challenge itself.

Over the past week, we've been making significant progress on Assembly, our experimental multi-agent AI workspace.

The idea behind Assembly is straightforward: instead of relying on a single AI model to answer a question, why not bring multiple AI agents together, let them reason independently, challenge one another's conclusions, examine evidence, and work towards a synthesized answer?

Building that experience into a usable, reliable product is a much bigger challenge. And that's exactly what we've been working on.

What's new in Assembly?

The last week has been particularly active, with several improvements making their way into the project.

1. A completely refreshed UI/UX

We've given the Assembly website and workspace a substantial visual and usability overhaul.

The landing page now has a more distinctive visual identity, an improved hero experience, refined navigation, and better content presentation.

The workspace has also received attention, with improvements to navigation, settings, interaction patterns, and the overall investigation experience.

A special thanks to the contributors who have been actively working on these improvements.

2. Better frontend reliability and accessibility

A visual redesign is only one part of building a good product.

We've also been working on frontend stability, navigation consistency, keyboard interactions, accessibility, and browser-level testing.

Some recent improvements include:

  • Fixes for navigation and history synchronization.
  • Improved accessibility and contrast.
  • More reliable interaction and dialog behavior.
  • Better automated browser-test coverage.
  • Improvements to investigation history and import handling.

3. Stronger foundations for an open-source project

As Assembly grows, maintainability and reliability matter just as much as new features.

We've continued improving automated checks, execution event schema validation, test coverage, and contributor-facing documentation.

The goal is to make Assembly easier to understand, develop, test, and extend.

What exactly is Assembly?

For anyone discovering the project for the first time, Assembly is an experimental multi-agent AI workspace.

You provide a question and an execution time budget. Assembly coordinates a temporary team of AI agents that:

  1. Plan the investigation and assign appropriate roles.
  2. Work independently on the problem.
  3. Compare answers, evidence, confidence, and disagreements.
  4. Challenge conclusions through debate, research, or adversarial checks where useful.
  5. Produce a synthesized response with sources and remaining uncertainty where available.

The project currently supports execution budgets of 3, 5, 10, and 15 minutes.

It follows a Bring Your Own Keys (BYOK) approach, allowing users to configure their own supported model-provider credentials.

One important distinction: Assembly is a research project, not a guarantee of correctness. Multiple models can still share blind spots, and important conclusions should always be independently verified.

We need more contributors!

Here's something I want to emphasize.

Assembly is a huge project, and there is still a lot to build.

Multi-agent orchestration, model integrations, evidence evaluation, backend reliability, frontend experience, testing, security, documentation, and desktop development are all substantial areas of work.

We don't expect a small team to solve everything.

That's why we're actively looking for more developers and open-source enthusiasts who would like to become part of the journey.

You don't need to be an expert in AI or multi-agent systems to contribute.

There are opportunities across:

  • Frontend development: React, Next.js, TypeScript, UI/UX and accessibility.
  • Backend development: Python, FastAPI, APIs and orchestration.
  • AI and integrations: Model providers, agent coordination, routing and evaluation.
  • Testing and reliability: Automated tests, debugging and performance.
  • Security and privacy: Credential management, application security and safe defaults.
  • Documentation: Developer guides, examples and technical writing.

Whether you want to work on a small issue, introduce a new feature, improve the interface, or explore the architecture, contributions are welcome.

Check the repository's open issues, especially those marked good first issue or help wanted, before getting started.

You can also use Assembly!

This isn't an invitation exclusively for developers.

If you're interested in experimenting with multi-agent AI, comparing how different models approach a problem, or simply exploring a different approach to AI-assisted research, you're welcome to try Assembly.

Your feedback is just as valuable as code contributions. Reports about confusing workflows, unexpected behavior, bugs, and feature ideas can help us make the project better.

Explore Assembly

GitHub Repository: https://github.com/anubhab7105/Assembly

Project Website: https://assembly-murex.vercel.app/

Check the repository for the latest setup instructions and project updates.

A Thank You to Everyone Contributing

I genuinely appreciate everyone who has already spent time reviewing issues, suggesting improvements, submitting pull requests, testing features, and helping shape Assembly.

Open-source projects grow through collaboration, and every meaningful contribution makes a difference.

We're still early in this journey, and there's plenty left to build.

Let's build something meaningful together.

If you're interested in contributing, trying Assembly, or sharing feedback, come say hello in the GitHub repository.

Happy building! 🚀

Top comments (0)