One workspace.
Multiple models.
Work with multiple AI models, compatible agents, and connected tools in one open-source workspace. Choose how you work with AI, on your terms.
Free to start. No credit card. v1.1.4
Cloud model · Requests go to your selected provider
At home on
your desktop.
Chat history stays on your device. Not on our servers.
Your models, together.
Connect the providers you use. Keep the workspace you love.
Local by design.
Yours by default.
Cosmo is an open-source, cross-platform workspace that keeps your conversations on your device and lets you choose how you connect to AI.
Set up your workspaceChat history lives on this device.
Local storage isn’t the same as local inference. Cloud prompts go to the provider you choose, under its privacy policy. Use a local model to keep prompts on your machine.
The full privacy pictureModels, agents, and tools. One place.
Bring together the AI models, agents, and tools you use for code, research, and everyday work.
Give your AI more context.
Connect MCP servers so your model can work with external tools and live data, not just the words in your prompt.
Bring the right information into the conversation.
Context added to chatMeet your local agents.
Install compatible ACP agents and give each one a dedicated workspace. Keep agent conversations right alongside your chats.
Stay in the flow.
Reach for slash commands without leaving your conversation. See model thinking, track tokens per message, and stop a response whenever you need.
Markdown, code highlighting, and a workspace that gets out of your way.
Next on the horizon.
In development. Not available yet.
- Web Search
- Workflow Automation
- Cost Intelligence
- Code Execution
Cost Intelligence in development
The real cost
is getting work done.
Developers can see what a model charges, but not what it costs to get work done. The cheapest model may need multiple attempts, or an agent can get stuck in a loop and consume resources without making progress.
Track cost by task.
See what it costs to get a piece of work done across models and agents.
Catch wasted spend.
Flag retries, redundant context and agent loops that waste resources without progress.
Choose the right model.
Get recommendations based on both cost and performance for the task at hand.
Get more done for less.
Help people get more work done with AI while spending less.
Start free.
Stay in control.
Bring your own keys, not another AI subscription. Pick the workspace that fits how you work.
A little clarity.
Cosmo is your workspace. Model usage is billed separately by your cloud provider. Local models run on your hardware.
Community
Freeforever
Community is the home for Cosmo’s complete AI workspace: models from multiple providers, local AI, MCP tools, ACP agents, personas, commands and much more
Download CommunityProfessional
Coming soon
Good questions.
Read the full FAQChats are stored on your device. When you use a cloud model, your prompt is sent to the provider you selected under its own privacy policy; local models can run on your machine.
Cloud providers may charge for API usage separately from Cosmo Studio. Running a model with Ollama or LM Studio avoids cloud API fees but uses your own hardware.
Connect providers such as OpenAI, Anthropic, Gemini, Groq, and DeepSeek, or run supported local models through Ollama or LM Studio. Choose a model when you start a chat.
Not yet, but it is on our roadmap. Currently, we support macOS, Windows, and Linux.
Not currently. Your chat history is stored locally on each device; do not rely on cross-device sync for backups or access to past chats.
Make room for
better work.
One desktop app. Your choice of AI.
A workspace that stays yours.
Free to start. No credit card. v1.1.4