Why Guild Stack
Serious AI work needs a route and a record.
Guild Stack is the workflow layer for the part after prompting: agree on the task, choose focused help, challenge the result, and preserve the evidence.
Page evidence
Origin and proof
A problem became an operating contract.
- 01 Problem
The hard part starts after the prompt.
Teams still need an agreed boundary, named decisions, and a result they can reopen.
Read the workflow contract - 02 Route
Approvals became visible stops.
Scope, team, and plan stay inspectable before specialist work begins.
Inspect the product map - 03 Record
The work leaves receipts in the project.
Run state, handoffs, review, verification, and provenance survive the chat.
Open the proof model - 04 Open source
The contract ships in public.
Guild Stack is MIT-licensed, with tagged releases and source available for inspection.
View the v2.0.0 release
Why it exists
About Guild Stack
Guild Stack is an open-source workflow layer for AI coding work inside your
existing codebase. It gives one task a visible route: approved scope, a focused
team, an agreed plan, review, verification, and local evidence under .guild/.
It was created by Lookatitude to solve a real problem: as AI-assisted development scales up, the hard part is not producing text or edits. The hard part is knowing what was agreed, what changed, what evidence exists, and what still needs human judgment. Guild Stack keeps those decisions and receipts connected so a team can reopen the work after the chat ends.
The first route is intentionally simple: run one bounded task and approve the scope, team, and plan. Experienced operators can then tune rigor, model tiers, and review strength. Harness developers can extend the same lifecycle to more hosts without hiding weaker capabilities.
Licence
Guild Stack is MIT-licensed. Contributions welcome via GitHub.
- Requestbounded task
- Specapproved intent
- Teamphase specialists
- Planlane sequence
- Buildhandoff receipts
- Reviewchallenge + rework
- Verifyfinal evidence
docs/v2/02-operating-model.md One system, three perspectives
Start with approvals. Add rigor, tiers, and review posture later.
Open source