madcompanyCommands
In the terminal
npx madcompany <command>
| Command | What it does |
|---|---|
init [--mode M] |
Set up madcompany in this repo: .madcompany/, the MCP server entry, the safety hook and the /mc-* skills. Never overwrites your files. M is ui, mvp, ui-mvp (default), brownfield or spec |
mode [M] |
Show how this project is built, its release ladder and its steps; or switch to M |
hq [--port 4317] [--share] |
Start HQ and the MCP server agents use. --share lets people you invite sign in from their own machines |
staff [--template small|medium|large] |
Generate one Claude Code subagent per team member; or start from a 5, 10 or 20-person company |
people add|link|remove|list |
Invite teammates (member) or clients (viewer) to HQ and manage their sign-in links |
scan [--json] |
Look at an existing codebase: stack, check commands, folders, services, env keys (brownfield) |
import [--epics F] |
Anchor BMad planning docs and turn stories into tickets (spec-driven) |
status |
One-screen summary: mode, current release, team, tickets, questions, open Human help |
merge <TICKET> |
The merge queue: merge an approved ticket into its release or epic branch, run the checks, then mark it done or revert and reopen it |
ship <R1|E1> |
Merge an approved release or epic into main, tag it and write release notes. Push and deploy go to Human help |
start-day, end-day, stop |
Workday controls (also buttons in HQ) |
snapshot |
Commit HQ's logs, facts and design docs, and nothing else |
env <agent|integration> |
Ports and database for an agent or the integration app |
refs check|fix |
Find, or link, bare references like "FR12" |
ref <ID> |
Everything linked to an ID: tickets, change requests, decisions, questions, notes, discussion |
trace [--strict] |
How much work and discussion maps to requirements, and what to fix. --strict exits non-zero for CI |
anchor <files…>, ids |
Add anchors to requirement IDs; rebuild the reference registry |
demo [dir] |
A sample project mid-way through a UI-MVP build |
In Claude Code
| Skill | What it does |
|---|---|
/mc-setup |
After installing as a BMad module: asks your mode and runs init |
/mc-staff |
Propose roles, domains and models for your app (or a template); generate the agents |
/mc-scan |
Brownfield: map the codebase, set the real checks, file Human help for missing keys |
/mc-plan-release |
Plan the next release with you: one goal, single-domain tickets, Human help filed early |
/mc-plan-epic |
Spec-driven: import BMad stories and split them into tickets |
/mc-ui-sprint |
Build and sign off screens with you; derive the API contract from them |
/mc-start |
Start the workday; your session becomes the lead and keeps the team busy |
/mc-check |
Check new stories and change requests against what's agreed; route each contradiction to the PM or architect |
/mc-review-feedback |
Plan a review you submitted: classify each comment (look, flow, feature), answer every conflict, propose one plan for your OK |
/mc-minutes |
Save the key points, decisions and actions of the current conversation to HQ |
/mc-end-day, /mc-stop |
End the day with handoffs, or stop immediately |
BMad skills each workflow uses
From BMad Method's current skill list. If a name differs in your version, /bmad-help shows the current ones.
| Skill | UI-driven | MVP-driven | UI-MVP | Brownfield | Spec-driven |
|---|---|---|---|---|---|
/bmad-product-brief |
✓ | ✓ | ✓ | ✓ | |
/bmad-spec |
optional | ✓ | |||
/bmad-prd |
✓ | ||||
/bmad-ux |
✓ | ✓ | ✓ | ||
/bmad-architecture |
optional | optional | optional | ✓ | |
/bmad-create-epics-and-stories |
✓ | ||||
/bmad-project-context |
optional |
In HQ
| Screen | What's there |
|---|---|
| Dashboard | What needs you, releases, who's working on what, blocked work, latest screenshots, decisions, today's activity |
| Chat | Channels, DMs and a thread per ticket. @mention anyone |
| Board | To do → In progress → Blocked → In review → Done, by agent, release or epic |
| Releases (Epics) | Each release's goal, tickets and progress; Approve or Request changes when it's yours to review |
| Library | Planning files, UX and mockups, architecture, the codebase map, release notes, meetings, conversations, links |
| Inbox | Multiple-choice questions for you, change requests, your saved answers |
| Human help | Everything only you can do, with steps and key checks |
| Tools | MCP servers, plugins and skills: who uses what, what agents may do without you, suggestions, recent calls |
| Decisions | What the team decided without you, and why |
| Trace | Coverage, what to fix, and every requirement with its tickets and change requests; a page per ID with everything linked to it |
| Team | Agents by department with their models and memory; hiring; the people who use HQ |
| Preview | Your running app at phone, tablet and desktop widths; comment on any element |
The agents' tools
Agents talk to HQ through an MCP server. The main tools: mc_claim, mc_block (park with a checkpoint), mc_submit, mc_review, mc_ask, mc_decide, mc_design_write, mc_human_help, mc_tools (which MCP servers and plugin skills are theirs), mc_ref and mc_section (look an ID up in the project index instead of reading files), mc_note, mc_conflict (raise a contradiction to the PM or architect), mc_memory_write, mc_handoff. The lead also has mc_release, mc_create_ticket, mc_escalate, mc_map_change, mc_requirement, mc_change, mc_withdraw_change and mc_check (with the PM), mc_trace, mc_ready_for_review and mc_wait, which waits for events without using any Claude usage.