The MySpec workflow is the platform's default methodology. Greenfield: a product interview, then constitution.md, requirements.md, solution.md, and tasks.md, each written through collect → generate → review. Brownfield: an interview that reads your codebase, then proposal.md, a coded requirements delta, and an optional task plan.
Day one they ask about the business, not the database. Day two they write down the rules everyone will build under. Then the brief, then the drawings, then the schedule - and at each step a second senior reviews the document before you sign it. That is the whole workflow; the AI plays both architect and reviewer, and you sign.
The four-file bundle
MySpec generate-workflow (greenfield) · myspec-greenfield-v1
- 1 Interviewno file yet
- 2 Constitutionwrites constitution.md
- 3 Requirementswrites requirements.md
- 4 Solutionwrites solution.md
- 5 Taskswrites tasks.md
- Completehand off via MCP or ZIP
Every file-writing stage runs three substages - collect (clarifying questions) → generate (draft the file) → review (an independent reviewer checks it against the template and your answers). You approve, edit, or send it back before the next stage starts.
Stage by stage
- Interview. A Product Interviewer asks one to five focused multiple-choice questions per round - the problem, product type, target users, the two or three key journeys, MVP scope - and deliberately nothing technical. Two or three rounds is typical. It ends by naming the session and the bundle (a kebab-case folder name).
- Constitution → constitution.md. Collect gathers technology, architecture, testing, coding-standard, and security constraints; generate writes the nine-section constitution; review checks every constraint traces to an answer.
- Requirements → requirements.md. A Business-Analyst collect loop keeps asking until the requirement list is complete (up to six rounds); generate writes FR-/NFR- entries with EARS+ criteria and requirement diagrams; review rejects untestable or technology-laden requirements.
- Solution → solution.md. Collect asks about the options the constitution left open; generate produces the architecture with flowchart, ER, sequence, and C4 diagrams; review checks constitution compliance and requirement coverage.
- Tasks → tasks.md. Collect asks about your team structure; generate writes milestones of sized tasks with estimates, a Gantt timeline, and a QA verification milestone; review checks every requirement has a task.
- Complete. The bundle is stored as versioned revisions and ready for hand-off.
The brownfield change flow
MySpec generate-brownfield · myspec-brownfield-v1
- 1 Interview (explores your code)no file yet
- 2 Proposalwrites proposal.md
- 3 Requirements deltawrites requirements.md
- 4 Taskswrites tasks.md optional
- Completehand off via MCP or ZIP
A much leaner sibling for existing code: no constitution or solution stage, because the codebase already embodies both. Instead:
- Interview that reads your code. The Brownfield Interviewer packs the repository (
local_pack_codebase), greps and reads the areas your answers point at, and asks only where a real gap exists. It cannot advance until real files were read - connect your code by mounting a folder (Chrome) or runningnpx -y @myspec/mcp-server reverse --root .. - Proposal → proposal.md: Why, What Changes, a Technical Solution with flow and schema diagrams, Impact - every claim grounded in files it read.
- Requirements delta → requirements.md coded AR- (add), BR- (remove, with
file:line), CR- (change, old criterion struck through). - Tasks (optional). After the delta is approved the workflow asks whether you want a task plan traced to the AR/BR/CR codes, or would rather hand the delta straight to your coding agent.
Folder layout
specs/
├── invoice-tracker/ # greenfield bundle (kebab-case name set in the interview)
│ ├── constitution.md
│ ├── requirements.md
│ ├── solution.md
│ └── tasks.md
└── add-oauth-login/ # brownfield change
├── proposal.md
├── requirements.md # Requirements Delta: AR-/BR-/CR-
└── tasks.md # optional
Spec paths are rooted at specs/. Downloaded through the MCP server they mirror into a gitignored .specs/ folder, so the copy your agent reads never collides with files you commit.
What the specialist passes add
- C4 diagrams in solution.md, when your organisation switches them on - a Context diagram of the system among its users and neighbours, and a Container diagram of the deployable units named exactly as the modules and infrastructure sections name them.
- Requirement diagrams in requirements.md - Mermaid
requirementDiagramblocks for FRs and NFRs, capped at 20 nodes each (25 on the brownfield workflow). - Project-Manager delivery plan in tasks.md - man-day estimates per task, totals in the overview, and a Gantt timeline against the team you described.
- QA verification milestone in tasks.md - one verification task per functional requirement, negative paths included.
- Skill recommendation - after the interview, up to six curated skills (framework, domain, or your organisation's custom skills) are suggested; the one you pick shapes every later file.
From bundle to code
The bundle is the contract your coding agent builds from. With the MySpec MCP server connected, the agent reads the files itself and writes its own updates back as reviewable revisions; without it, download the ZIP and drop the files into .specs/.
# Cursor / Claude Code, with the MySpec MCP server connected
"List my projects, then download every spec file of 'invoice-tracker' into .specs/."
"Implement task 3 from .specs/invoice-tracker/tasks.md strictly according to
solution.md and constitution.md. Cite the FR ids in the PR description." Need to change something later? The Edit workflow (greenfield or brownfield) takes the files you reference, your description of the change, and proposes each edit as a diff you accept, refine, or discard - keeping the four files coherent with each other.
Other workflows
The default four-file bundle: interview → constitution → requirements → solution → tasks.
GitHub Spec Kit's feature-scoped flow: interview → constitution → specify → plan → tasks.
OpenSpec's change-proposal flow: explore → proposal → delta specs → design → tasks.
Files this workflow writes
Project vision, core principles, and the MUST / MUST NOT constraints every later file has to obey.
User roles, FR-/NFR- requirements with EARS+ acceptance criteria; a coded AR-/BR-/CR- delta for brownfield changes.
Architecture, modules, data model, and API design with Mermaid flowchart, ER and sequence diagrams - plus C4 diagrams when your organisation enables them.
Ordered, sized implementation tasks grouped into milestones, traced back to requirement IDs.
Why a brownfield change is needed, what changes, the technical solution, and its impact on existing code.
Spec Kit's prioritised user stories and success criteria, or OpenSpec's ADDED / MODIFIED / REMOVED / RENAMED requirement deltas.
Spec Kit's technical plan: technical context, constitution gates, architecture, data model, and key decisions.
OpenSpec's optional context, goals / non-goals, decisions, and risks for a change.
Let the AI Architect write it for you
MySpec interviews you, then drafts every spec file with collect → generate → review gates. Free during Open Beta.