Skip to Content
Getting started

Getting started

Conclave runs on Claude Code or Cursor against the same conclave/ directory.

Two commands to get going

Inside any git repo:

# 1. One-time wizard — detects stack, collects project name, story prefix, launch date,
#    team profile, and finds the product document (docs/mvp.md, docs/project.md, etc.)
/conclave-init
 
# 2. Generate stories + architecture from the product doc, then lock Sprint 1 active.
#    Use --all to plan every sprint from the document in one pass.
/conclave-planning
/conclave-planning --all

That’s it. After step 2 the sprint is active, every story has an assignee and a discipline, and the team can open a PR with the entire conclave/ directory.

/conclave-init is a pure wizard (no AI agents). It auto-detects your stack, then asks for: project name, story ID prefix (e.g. US, TASK, FEAT), project language, launch date, team mode (solo/team), and team profile. It also locates your product planning document — or helps you create one.

/conclave-planning does the rest. On the first run it invokes the Tech Lead and Product Manager in parallel — reading your product document — to produce the backlog, architecture, per-story files, and Gherkin acceptance criteria. Then (on every run) it runs the Sprint Planning ceremony.

What gets created

After running both commands, your repo has a complete conclave/ directory:

conclave/
├── README.md
├── config.md                         # story_prefix, product_doc_path, team config
├── team/
│   ├── roster.md                     # who covers which discipline, plus optional PM/SM
│   └── ceremonies.md
├── product/
│   ├── backlog.md                    # ordered Product Backlog (from /conclave-planning)
│   ├── architecture.md               # ADRs + tech decisions (from /conclave-planning)
│   ├── definition-of-ready.md
│   └── definition-of-done.md
├── context/                          # frozen snapshots of inputs used
└── sprints/
    └── SPRINT-001/
        ├── meta.md
        ├── spec.md                   # sprint plan
        ├── planning.md               # planning meeting record
        ├── stories/                  # one file per user story (PREFIX-001-slug.md)
        └── acceptance/               # Gherkin acceptance criteria (AC-PREFIX-001.md)

Every file is markdown. Every change is a normal git diff your team can review in a PR.

Continuing the loop

Once Sprint 1 is locked:

# Each dev picks up an assigned story
/conclave-dev US-001
 
# QA verifies the story when it reaches review
/conclave-qa US-001
 
# Tech Lead approves the PR (only when peer_pr_review.required is on)
/conclave-pr-review US-001

When all the sprint’s stories are done, you’re ready for the next planning cycle. Run /conclave-planning for SPRINT-002 — Phase A is skipped (backlog already exists), only the planning ceremony runs.

Next steps

Last updated on
IntroductionInstallation