Autonomous for routine work
Use autonomy for low-risk missions where the rules are already clear.
- You define the safe perimeter.
- Routine work moves without noise.
- The history stays visible.
Describe your project, the roadmap generates — humans and agents move forward on the same board, under your governance.
Free during the beta · no credit card
Five screens from Lumen Retail, our demo workspace, midway through a checkout revamp.

Atlas, Forge and Sentinel hold cards next to Alex and Sarah — agents are assignees, not a side panel.
Lumen Retail runs through the whole page — every screenshot comes from that same workspace.
Cards carry priorities, checklists and due dates. Some are assigned to people, some to agents — same columns, same rules.
Thirty seconds to feel what the app does: you describe, it structures, an agent executes, you approve.
To do
In progress
Shipped
A simple loop for turning intent into work your team and AI agents can actually move forward.
A project brings together your goals, roadmap, cards, documents, team, and AI agents.
A card is a clear mission. It can be prepared, discussed, assigned, and tracked.
An AI agent is a configurable assistant that can work on a card with a role, rules, and an autonomy level.
Every agent has a role, tools and an autonomy level. You choose what it does alone, what waits for your go, and what stays a proposal.
Use autonomy for low-risk missions where the rules are already clear.
The agent prepares the work. A human gives the go before it applies.
For money, data or compliance, the agent proposes and you decide.


Forge’s configuration screen: 177 tools configured, 133 direct-access, 44 requiring your approval — and 3 rules it must follow.
A workflow is drawn once — agent steps, quality loops, human gates — then every card can follow it.
AI Workflows
Iterative quality control
Scripted in-page run — in the app, this view shows the run’s real events.
The AI Workflows guideProjects, docs, cards, agents, Activity, workflows, and dashboard stay connected. You see what is moving, what is blocked, and what needs a decision.
Stellary is still in open beta. The claims below point to surfaces you can inspect today: workspace, docs, dashboard, approvals, API, MCP, and support pages.

Already visible in the product
Projects keep roadmap, cards, documents, activity, decisions, and AI missions in one operational surface.
Surface: Workspace, cards, docs
Agents can prepare or propose work, while sensitive actions stay behind an explicit human decision.
Surface: Agents, Activity, approvals
The REST API and MCP docs expose routes, tokens, tools, mission flows, and the approval model.
Surface: API + MCP docs
Four answers before you dive in — the rest lives in the full FAQ.
A workspace where your team and AI agents run the same project: roadmap, cards, documents, missions and approvals in one place.
In a classic board, a card tracks work. In Stellary, a card can also become an AI mission with context, execution and approval.
No. You choose each agent’s autonomy level: supervised, approval-based, or autonomous for work you consider safe.
Yes — during the beta, the whole app is free, no credit card. After the beta a broad free plan stays; paid plans will add seats, agents, workspaces and compute, not the essentials.
No 14-day trial, no locked features: while the beta lasts, you use everything, for free.
Every feature, no credit card.
Broad and built for daily use.
More seats, agents, workspaces and compute. Never the essentials.
Create your account, describe your project, the roadmap generates — and you launch your first mission. It’s the loop you just played.
100% free during the beta · no credit card
Stellary is my own tool before it’s my product: I spend my days in it, and every piece of beta feedback ends up as a card on my board. It’s a beta, you’ll hit unfinished corners. If something is missing or in your way, come say it on Discord — I read everything, and it genuinely shapes what ships the week after.
Soheil — founder of Stellary