Turn project work into clear cards that people and AI agents can prepare, run, review, and approve in one shared board.
Launch workspace
RunningTo do
12Prepare release mission
Project context
In progress
12Implement MCP access
Allowed tools
Review
12Review agent output
Approval policy
A card carries the mission, context, assignee, execution state, questions, result, and approval history from start to finish.
01
Define the outcome, owner, priority, and acceptance criteria.
02
Attach project documents, links, decisions, and constraints.
03
Assign a workspace agent or an external runtime to the mission.
04
Read Activity, answer questions, and inspect the result.
05
Approve the work and keep its complete trace on the card.
Mission STL-128
Define the outcome, owner, priority, and acceptance criteria.
Checklist
Linked context
Cards, documents, comments, decisions, and project rules stay available to both people and agents.
Humans define the mission. Agents work inside the same project context and leave a visible trail instead of operating in detached chats.
Cards, documents, comments, decisions, and project rules stay available to both people and agents.
Questions, tool calls, progress, errors, proposals, and results return to the card Activity.
Choose whether an agent acts directly, works under supervision, or proposes writes for approval.
{
"mcpServers": {
"stellary": {
"url": "https://api.stellary.co/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN"
}
}
}
}| Action | Human | Supervised | Approval |
|---|---|---|---|
| Read cards and documents | Direct | Direct | Direct |
| Update a card | Direct | Policy | Proposal |
| Run a mission | Direct | Policy | Proposal |
| Use connected tools | Permission | Allowlist | Proposal |
A traditional Kanban card records work. An agent-ready card also provides a bounded mission: what must be produced, which context may be used, who or what is responsible, and how the result will be reviewed.
The agent does not need a parallel planning system. Its questions and output return to the same card the team already follows.
Agent activity should not turn the board into a stream of technical logs. Stellary keeps the card status, owner, priority, and outcome readable while detailed execution remains available in Activity.
The result is a shared operational view: people see what is ready, running, blocked, waiting for review, or complete.
Traditional Kanban
Agent-ready Kanban
Operators
Traditional Kanban
Designed mainly for human updates.
Agent-ready Kanban
Designed for humans and AI agents working together.
Context
Traditional Kanban
Often split across chats, docs, and memory.
Agent-ready Kanban
Structured on the project and attached to the mission.
Execution
Traditional Kanban
Work happens outside the card.
Agent-ready Kanban
The card can launch and follow an agent mission.
Control
Traditional Kanban
Automation is separate and difficult to audit.
Agent-ready Kanban
Permissions, approvals, and Activity are part of the workflow.
An AI Kanban board is a visual workflow where AI can help prepare, execute, update, or review cards. In Stellary, agents work from the card’s project context and return their progress and result to Activity.
Yes. They can run as local Stellary runners or connect through the Stellary MCP server. Their available actions still depend on the user or agent identity, project access, tool allowlist, and autonomy policy.
Only when you configure them to do so. Approval mode turns non-read actions into proposals. Supervised mode allows safe actions while protected tools still require review. Autonomous mode remains bounded by permissions and tool policy.
No. Documents remain the durable context for briefs, architecture, decisions, and requirements. Cards link that context to a precise piece of work and its execution history.
Start with one project, one board, and one controlled mission.
Create your first project