Explore the Stella Loop method

Build toward your
North Star.

Stella Loop is agent-native project management. Point every human and coding agent at a shared aim — and see, after each loop, whether the product actually got better.

A product system with a direction, a loop, and a measure.

Stella Loop starts before the ticket and learns after the merge — so you always know why work exists, and whether it made the product better.

Fig 0.1
Pulled by your North Star Three principle documents converge on a composite North Star, which remains visibly connected to active work.

Pulled by your North Star

Turn product principles into live inputs that guide analysis, proposals, and priorities — not passive documentation.

Fig 0.2
Built around the loop An outer product loop contains smaller implementation, fix, and re-specification loops.

Built around the loop

Keep discovery, planning, implementation, and review connected — including the fix and re-spec paths that real development requires.

Fig 0.3
Progress you can prove Successive product states rise along a measured trajectory, with score changes and area-level deltas. 71+5 76+5 82+6 86+4

Progress you can prove

Measure quality, track meaningful deltas, and see each cycle move the product forward.

Direction

Make product direction operational.

Describe what your product should become — that's your North Star. Analysis scores against it, proposals cite it, and every piece of work stays connected to why it exists.

Signals

Give every observation a front door.

A GitHub issue, a Slack thread, a Sentry alert, a passing thought — capture it in seconds, triage it once, and nothing the world tells you gets silently lost.

Discovery

See what the product needs next.

Analyzers examine your actual product — user experience, code quality, security, gaps — and return scored reports on what to improve. See the product, not just the backlog.

Decisions

Explore before you commit.

Findings become proposals that compete in one prioritized pool. Promotion is the commitment point — the only way an Epic is born — so you explore cheaply and commit deliberately.

Delivery

Carry context all the way to code.

Every task carries its context: the signal, the spec, the aim. Humans and agents work from the same board, and approval gates keep you in charge of what ships.

Improvement

Make every release start the next loop.

Review can send work back for fixes or a rethink — then the merge updates your scores, learnings carry forward, and the next loop starts smarter.

Opinionated about the loop. Open about the tools.

Stella Loop is the system of record for product work — it takes the place of your issue tracker, and connects to everything else. Feedback channels, spec tools, models, and version control all plug into one loop.

Signals in GitHub Issues, Slack, Sentry, inbound email, and signed webhooks land as triageable signals.
Spec tools OpenSpec by default; bring the planning workflow your team already trusts behind one adapter.
Analyzers Agent-backed analysis and deterministic tools, metrics, and checks under one manifest.
Models & budgets Route each stage to a model tier — frontier, standard, economy — with visible costs and hard stops.
Version control A GitHub App coordinates one repository or many; git by default, Jujutsu supported.
Agent surfaces The stella CLI, the public API, a generated MCP server, webhooks, and a realtime event feed.

One loop. Your stack.

For solo builders and full product teams.

One developer with agents is a product operation now. The same loop scales from a side project to an organization.

Vibe coders Your app works. Now make it excellent. Point Stella Loop at the app you built with AI. Analyzers find what needs attention, agents fix it, and scores prove it got better.
Solo developers Automate the development loop. Standing intents keep analysis running, agents claim ready tasks, and checkpoint gates only ask for you where it matters.
Product teams Adopt agents without losing control. Approval gates, budgets, and review panels give organizations real oversight — from fully supervised to fully autonomous, one stage at a time.

Set the aim. Start the loop.

Give every human and agent a shared direction — and a system that can prove the product is moving toward it.

Better products, one loop at a time.