Questions about Stellary

Stellary is an AI-native project piloting platform for technical teams. It connects projects, cards, documentation, agents, missions, pipelines, cockpit, and governance surfaces inside the same system.

Primarily product and engineering teams, CTOs, technical founders, and organizations that want humans and AI agents to work inside the same workspace with more control and less context loss.

No. The delivery and context core already works through projects, views, cards, documents, roadmap, and collaboration. The AI layer augments that system rather than replacing it.

Stellary supports multiple project entry points: quick creation, assisted creation, and project import. Startup initializes a delivery foundation with a scope and base views, then the workspace becomes the operating frame for members, agents, and documents.

The workspace is the main operating unit. It brings together members, teams, agents, knowledge base, projects, cockpit, plugins, skills, pipelines, and AI configuration. Above it, the organization carries cross-workspace governance.

A project connects scopes, views, cards, roadmap, documents, and project members. It is not just a ticket container; it is the concrete delivery unit of the product.

The delivery core currently covers board, list, calendar, timeline, and stats. A single project can therefore be read from multiple angles without losing the real work state.

A card can concentrate status, priority, dates, human or agent assignment, checklist, subtasks, dependencies, labels, custom fields, comments, attachments, linked documents, and even an agent mission.

No. Stellary includes workspace knowledge, project documents, and documents linked to cards. This documentation layer frames, transmits, stores, and directly feeds both human and agent execution.

The board shows work in motion. The cockpit turns that delivery into a piloting view: active missions, blockers, arbitrations, live activity, health signals, and briefings. It is built to govern execution, not to duplicate cards.

An agent is a software operator attached to the workspace. It has a role, status, autonomy mode, AI profile, allowed tools, rules, skills, and project assignments.

Three modes structure the product today: supervised, approval, and autonomous. That lets you choose between controlled review, proposal validation, and more direct execution depending on trust level.

Depending on its framing, an agent can load a card context, read comments and linked documents, execute its mission, publish a result, update the real card status, and surface signals back to the cockpit.

The agent runtime uses the current mission, run history, agent, team, or workspace memory, project documents, and card-linked documents. Agents do not restart from zero on every execution.

Stellary MCP exposes the same tool graph, permissions, and control rules as the in-app AI runtime. A compatible client can therefore read or act on the real product context without duplicating business logic somewhere else.

Yes. Stellary tracks missions, runs, pending proposals, costs, token volumes, average duration, tool history, and latest activity. The goal is to make AI usable and governable in production.

Yes. The real hierarchy is Organization -> Workspace -> Project. The organization carries plan, usage, billing, and workspaces; the workspace carries day-to-day operations.

Yes. They are native workspace building blocks. Plugins open external capabilities, skills specialize agents, and pipelines orchestrate repeatable sequences between agents, human validation, and notifications.

The codebase already exposes a workspace plugin catalog including GitHub, Email, Slack, Discord, and an example plugin, on top of the MCP server and internal tools available to the agent runtime.

The product covers sign-up, sign-in, email verification, forgot password, reset, password change, TOTP 2FA, personal API tokens, and account deletion. Workspaces and organizations then apply their own roles and permissions.

Yes. Stellary exposes user, workspace, and organization settings, along with a super admin layer. Current surfaces cover plan, subscription, usage, billing, members, and AI configuration.

No. Humans go through the application auth flows. Agents are service users governed inside the workspace, with their own tokens, permissions, and autonomy modes.

No. The repository is currently distributed under a proprietary license. The product may have a GitHub presence and public documentation, but this FAQ does not present it as an open-source project.

Yes. The repository separates the backend, the main frontend, and the website, with environment examples, PostgreSQL as a prerequisite, and helpers such as docker-compose for a local stack.

Backend uses NestJS with Prisma and PostgreSQL, the product frontend uses React + Vite + Tailwind CSS, the website uses Next.js, realtime uses WebSocket / Socket.IO, AI integration relies on OpenAI SDK and MCP, and the codebase is predominantly TypeScript.

The canonical entry point is docs/README.md. Product documentation is organized there by vision, platform map, functional domains, and technical references.

The website currently points to Stellary GitHub for product and technical questions. Repository documentation remains the best starting point if you want to understand the product as it exists today.