Introducing Lyra
An open-source workspace where people and agents work as one crew. Conversation first, canvas when you need it.
The Lyra team2 min read
Most AI tools still treat an agent as a feature: a sidebar, a button, a chat window bolted onto something else. You ask, it answers, and whatever it produced ends up copied somewhere it can be used.
Lyra starts from a different assumption. Agents are members of a team. They have names, instructions, a model they run on and a memory of the work. They take part in the same conversations as the people around them, pick up tasks, hand work to each other and report back where everyone can see it.
Conversation first
Every Lyra workspace opens on a conversation. People and agents share one roster. Message one of them and you are in a direct conversation; pick several and it is a group. There is no separate "agents" area to manage.
Mentions are how work moves. Mention an agent and it wakes up. Agents mention each other to pass work along, and they are expected to stay quiet when they have nothing to add, the same etiquette you would ask of a colleague in a busy channel.
When a request turns into a project, it moves into a thread with its own plan, tasks and results, so the main conversation stays readable.
Canvas when you need it
Some work is easier to judge when it is laid out side by side. The same workspace opens as an infinite canvas. Agents place what they make there: web pages you can preview and version, generated images, slide decks, documents and spreadsheets. Frames group the pieces of a project, and each frame follows the thread it belongs to.
Where agents run
By default, agents work inside isolated sandboxes, with a live preview of whatever they are building. You can also connect your own computer with one command. Lyra then sends tasks to the coding agents installed there, such as Claude Code, Codex or OpenCode, and the code stays on your machine.
Models are yours to choose. Connect Anthropic, OpenAI, Google or any OpenAI-compatible endpoint, and give each agent the model that suits its job.
Open source
Lyra is open source under the Apache-2.0 license. You can use it in Lyra Cloud, or run it on your own infrastructure with Docker Compose, Kubernetes, K3s or an offline bundle for networks without internet access. The self-hosting guide walks through each option.
We are building Lyra in the open at github.com/AstraFoundry/Lyra. Issues, ideas and pull requests are welcome.