01 / CONTEXT
Give every project
a shared memory.
Files, instructions and decisions travel with the work. Agents start with the context you select and return answers linked to their sources.
Explore projects →A PRIVATE AI WORKSPACE
Your work.
Your agents.
Your power.
Bring files, local AI models and everyday work together. Give agents a task, follow their progress and keep the results in your project.
For individuals and small teams. On a computer or private server you control.
Studio / Projects
Local models
PROJECT / PRODUCT LAUNCH
Launch brief
A clear plan, grounded in your project files.
Prepare the launch briefReady for review
- ✓Reader · collect source materialComplete
- ✓Analyst · compare prioritiesComplete
- ✓Editor · prepare the briefComplete
launch-brief.md
Draft ready. Two decisions need your review before sharing.
positioning.pdf · p. 4roadmap.md · §2Project sourcesRead access
▤ positioning.pdfPDF · 8 pages
▤ roadmap.mdMarkdown
▤ interviews.txtText · 6 notes
Context stays with the project.
Sources, drafts and accepted outputs have separate places.
Project agentsOn your host
Reader
Find source passages. Read project files.
Analyst
Compare findings. Write working notes.
Editor
Prepare a draft. Request approval to publish.
A role is a set of instructions.
Agents can share one local model or use different models for different tasks.
Product tour · example project
Your files · one project contextYour models · local inferenceYour decisions · review before action
A WORKSPACE THAT DOES THE WORK
From scattered information
to something you can use.
Keep the brief, the source material, the people and the work in the same place.
Give every project
02 / DELEGATION
A capable team.
A clear set of boundaries.
Readers, analysts and editors work toward the same outcome. You decide what each agent can read, write and ask to do.
Meet your agents →03 / REPEATABILITY
Good work deserves
a repeatable process.
Turn a useful sequence into a workflow. Define its inputs, review points and outputs, then run it again as your project changes.
Explore workflows →COMPUTE CLOSE TO YOUR WORK
A powerful workspace.
On your terms.
Run open-weight models on your workspace host. Keep document preparation, retrieval and model inference close to the files they use.
Choose the models that fit your hardware and the work. Set how much runs at once. Keep a useful model warm instead of loading it for every task.
Understand local deployment →Your computer or private server
Deskarden workspaceProjects · files · people · review
Agent workersInstructions + allowed tools
Project contextSources + saved outputs
Local model runtimeOpen-weight models on your hardware
Inference stays on your host. External research and connectors are separate permissions.
START WITH SOMETHING USEFUL
Everyday work.
A little more capable.
Build a grounded brief.
Read project material, compare findings and prepare a brief with source references and open questions.
See the workflow →Give documents a home.
Extract key details, suggest a destination and stage a tidy project record for your review.
See the workflow →Turn activity into an update.
Bring together recent work, decisions and blockers. Create a concise update without rebuilding the context by hand.
See the workflow →HOW IT COMES TOGETHER
One place to think.
One place to finish.
Each run leaves a visible trail: the sources it used, the steps it completed and the output you chose to keep.
Read the workspace guide →Bring in the context
Create a project, add its files and write down the goal.
Choose the workflow
Pick an agent or a saved sequence. Set its access and review points.
Review and keep the result
Check the sources, resolve open questions and accept the output into your project.
GOOD QUESTIONS
A few things to know.
Where do the AI models run?
On the computer or private server that hosts your workspace. Your browser is the interface. Local inference does not require sending project content to a separate hosted model provider. Read the deployment guide for the boundaries.
Can I use it on an ordinary computer?
The model, its memory requirements and the size of your tasks determine what fits. Smaller local models can suit modest workloads; larger models and simultaneous runs need more memory and compute. You choose the runtime and resources for your deployment.
What makes an agent different from a chat?
An agent has a role, selected context and permission to use particular tools. A workflow connects those tasks into a visible sequence with saved outputs and review points. Chat remains useful for shaping the task and asking follow-up questions.
Does anything leave my workspace?
Local file processing and local inference stay on the workspace host. If you enable web research, an external connector or an action that sends information, that operation crosses the boundary. Permissions should reflect that choice. See security and data boundaries.
How do I get started?
If your workspace is already set up, sign in using your account email. For a new deployment, contact us with your intended use and hardware.
Make room for a better way to work.
Bring your projects, models and workflows into one private workspace.