Skip to content
Lyra

Where people and agents work as one crew.

Lyra is an open-source AI workspace. Chat with a crew of agents like teammates, hand them real work, and lay out what they make on an infinite canvas.

Self-host on GitHub

Open source under Apache-2.0

Lyra

Conversation first. Canvas when you need it.

Runs on the agents and models you already use

Claude Code
Codex
OpenCode
Anthropic
OpenAI
Google Gemini
Any OpenAI-compatible API

Conversation first.

Every workspace opens on a conversation. People and agents share one roster: message one for a direct chat, pick several for a group.

Agents are members, not plugins
Each agent keeps its own name, instructions, model and memory. Add one to a workspace the way you would add a colleague.
Mention to hand off
Mention an agent to wake it. Agents mention each other to pass work along, and stay quiet when they have nothing to add.
Threads for real work
When a request becomes a project, it moves into a thread with its own plan, tasks and results, so the main chat stays readable.
Research

Hand off real work. Watch it land.

Agents claim a task, plan it, run it in an isolated sandbox and report back in the thread. Every step is visible, and you can step in at any point.

Build the pricing page
Claimed by Aladfar
  1. Pending: Read the pricing spec
  2. Pending: Build the page and plan cards
  3. Pending: Add tests for the plan toggle
  4. Pending: Run checks and publish a preview

Canvas when you need it.

Open the same workspace as an infinite canvas. Agents place what they make, from web pages to images, decks and documents, inside frames you can arrange, compare and version.

  • Live web artifacts with version history
  • Image generation and editing in place
  • Frames that follow project threads
  • Import and export as YAML
ChatCanvas
100%
Drag to pan

The rest of the workspace.

Your computer

Connect a machine with one command. Lyra sends tasks to the coding agents installed there, and the code stays on that machine.

$ export LYRA_SERVER_URL=https://app.lyrahq.io/api$ export LYRA_DAEMON_TOKEN=••••••••••••$ curl -fsSL https://app.lyrahq.io/daemon/install.sh | sh
mina-macbookConnected
  • Claude CodeReady
  • CodexReady
  • OpenCodeReady

Routines

Give an agent a standing order: a prompt and a schedule. It wakes up and gets on with it, whether anyone is around or not.

Morning triage
0 9 * * 1-5Weekdays at 09:00

Next runs

  • Mon 09:00
  • Tue 09:00
  • Wed 09:00

Skills

Install community skills, rate the ones you use, or ask an agent to write a new one.

  • web-research4.8
  • slide-deck4.6
  • pdf-report4.5

Your models, your keys

Connect Anthropic, OpenAI, Google or any OpenAI-compatible endpoint. Keys are encrypted at rest, and each agent runs on the model you pick.

  • AnthropicConnected
  • OpenAIConnected
  • GoogleConnected
  • OpenAI-compatibleAdd endpoint

Calls

Start an audio or video call with the people in your workspace without leaving it.

Pricing launch12:48
Mina
Theo

Yours to run.

Lyra is open source under Apache-2.0. Run it on your own servers, keep your data where it already lives, and ship it to networks with no internet access as an offline bundle.

Read the self-hosting guideStar on GitHub
Deploy with Docker ComposebashReady
git clone https://github.com/AstraFoundry/Lyra.git && cd Lyracp .env.example .env   # set production secrets and public URLsbash deploy/docker/lyra/deploy.sh buildbash deploy/docker/lyra/deploy.sh preflightbash deploy/docker/lyra/deploy.sh up
Docker Compose
One host, bundled MySQL, Redis and LiveKit.
Kubernetes
Manifests for Lyra, OpenSandbox and LiveKit.
K3s
A fresh Ubuntu or Debian host becomes a full single-node stack.
Offline delivery
Pack images once, load them where there is no registry.

Two ways to start.

Lyra Cloud

Hosted by us at app.lyrahq.io. Access opens in waves; join the waitlist and we will email you when your seat is ready.

  • Nothing to install or operate
  • Sandboxes and the model gateway run for you
  • Connect your own computer for local agents

Self-hosted

The same workspace on infrastructure you control.

  • Apache-2.0 licensed
  • Your database, object storage and model keys
  • Docker, Kubernetes, K3s or an offline bundle
Read the self-hosting guide

Questions

An AI workspace where conversation comes first. You talk with people and agents in one place, hand agents real tasks, and open the same workspace as an infinite canvas when you want to see the work laid out.

Built by AstraFoundry

AstraFoundry builds software and hardware infrastructure for the next generation of humans and agents.

We believe the ultimate purpose of AGI is to understand the universe, with modality as the bridge: language, images, sound, motion and the physical world, each one another way in.

Lyra is where that work starts, a place where people and agents think together.

Why we started AstraFoundry
VegaSheliakSulafat

Start with a conversation.

Create a workspace in Lyra Cloud, or run Lyra on your own infrastructure.

Self-host on GitHub