Structured Workflow MCP
This server enforces disciplined, phase-based development workflows for AI assistants, requiring structured documentation and validation at each stage before proceeding.
Workflow Types:
Refactoring workflows - Improve existing code through mandatory audit, analysis, planning, implementation, testing, and iteration phases
Feature development workflows - Build new functionality with integrated testing and documentation requirements
Test coverage workflows - Add or improve test coverage with systematic analysis
TDD workflows - Implement features using Red-Green-Refactor cycles with test-first principles
Custom workflows - Design workflows with full control over phases, iteration limits, and output preferences
Core Capabilities:
Enforce sequential execution - Mandate specific phases in order, preventing skipping through session state management and validation
Require verified outputs - Force production of structured artifacts (JSON, markdown) at each phase completion using the
phase_outputtoolValidate actions and progress - Check safety rules (e.g., files must be read before modification via
validate_action) and verify phase completion meets requirementsProvide phase-specific guidance - Deliver focused instructions for each phase (SETUP, AUDIT_INVENTORY, COMPARE_ANALYZE, QUESTION_DETERMINE, WRITE_OR_REFACTOR, TEST, LINT, ITERATE, PRESENT)
Track workflow status - Monitor current progress and session state throughout execution
Handle escalation - Trigger user input when iteration limits are reached, checkpoints are hit, or validation failures occur
Auto-suggest file organization - Generate numbered, timestamped file naming for workflow documentation (e.g.,
01-audit-inventory-2025-01-04.md)
Configuration Options:
Set iteration limits for LINT, TEST, and ITERATE phases
Define user checkpoints for major changes or after failures
Customize output formats (markdown/JSON), directories, real-time updates, diagrams, and code snippets
Scope workflows to specific files, directories, or entire projects
Includes a 'Buy Me a Coffee' link in the README for supporting the developer through donations.
Built with TypeScript, providing type-safe development for the MCP server implementation.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Structured Workflow MCPrefactor the authentication module to improve security"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Structured Workflow
Work in progress. The 20 workflow skills are available for review and experimentation. Project initialization, templates, companion subagents, hooks, and packaged installation are still being designed.
Structured Workflow started as an MCP server in the Sonnet 3.5/3.7 era, when I was trying to answer a question: “How do I get my agents to follow a more structured approach, like a human would, when facing new problems and unknown territory?” This update evolves that idea by bringing in the MYP Design Cycle from my years as a teacher as the core process for building anything with agents.
Structured Workflow is a collaborative skillset for agents and humans. It gives coding agents an evidence-producing design cycle as a workflow structure, guiding them through the reasoned process that a product developer, designer, or engineer would use when approaching and solving a problem. It can be lightweight or support detailed analysis and research. It provides a useful default direction while allowing the work to follow new questions, ideas, prototypes, implementation discoveries, and evaluation evidence wherever they lead.
The repository currently contains 20 skills: one parent and four criterion skills for each of four phases. It also defines the durable project files that preserve shared understanding and workflow position across sessions.
The design cycle
The default direction is to work through phases A–D, one criterion at a time. Work can move to any criterion or phase that best answers the current question. Research, ideas, exploratory visuals, and evaluation can move naturally between criteria as the current question changes.
A — Inquire and Analyze
A1 — Explain and justify the need
A2 — Identify and prioritize research
A3 — Analyze prior art
A4 — Develop the Design Brief
The phase produces an approved Design Brief supported by inquiry evidence.
B — Develop Ideas
B1 — Develop design specifications
B2 — Develop feasible ideas
B3 — Present and justify the chosen design
B4 — Develop planning drawings and diagrams
The phase produces an approved Spec that defines the chosen solution.
C — Create Solution
C1 — Construct a logical plan
C2 — Demonstrate technical skills
C3 — Follow the plan to create the solution
C4 — Justify changes to the design
The phase produces approved tickets, a working solution, verification evidence, and justified changes.
D — Evaluate
D1 — Design testing methods
D2 — Evaluate the solution against the design specification
D3 — Explain how the solution could be improved
D4 — Explain the solution's impact
The phase produces an evidence-backed verdict, improvements, and an impact explanation.
The default direction inside A and B follows criterion order. Criterion C starts with C1, demonstrates C2 while carrying out C3, and uses C4 when creation requires a design change. A complete final evaluation follows D1 through D4.
Evaluation is also available throughout the cycle. A Design Brief, Spec, plan, ticket set, prototype, implementation, or claim can be evaluated whenever the work needs an evidence-based judgment.
Related MCP server: Universal Spec Architect
How artifacts develop
Each phase maintains one authoritative project document:
The inquiry document preserves A1–A4 evidence and produces the Design Brief.
The developing-ideas document preserves B1–B4 evidence and completes the Spec.
The creating-solution document preserves the logical plan, technical skills evidence, creation record, and justified changes. Approved implementation tickets live in the project's issue tracker.
The evaluation document accumulates reviews and final evaluation evidence.
The documents preserve the reasoning behind each handoff. A later phase can load the approved artifact first and retrieve deeper evidence from the owning phase document when needed.
Durable project context
Structured Workflow uses a small set of project files to survive context loss:
AGENTS.mdcontains the short, always-loaded workflow rules.workflow-tracker.mdrecords the current phase, exact criterion, current question, active artifact, target result, and likely next move.GLOSSARY.mdpreserves the shared meaning of important project terms.Phase documents preserve findings, evidence, decisions, and uncertainty.
Project templates define the document locations and reusable expectations, including the project's Definition of Done.
The tracker records position rather than findings. Research and reasoning belong in the phase document that owns them. Implementation work belongs in the issue tracker.
Human-agent collaboration
Prefer retrieval-led reasoning over training-led reasoning. The agent begins with the project, its code or materials, direct human knowledge, and primary sources. Material claims are cited where they are recorded. Supported facts, human judgments, inferences, and unknowns remain distinguishable.
The agent develops questions from the request, project evidence, industry evidence, contradictions, constraints, and missing understanding. It leads retrieval, review, synthesis, and other bounded work. The human directs intent, priorities, values, and consequential choices. The balance can be even or mostly agent-led, depending on the human and the work.
This evidence trail gives the human a concrete basis for inspecting the agent's reasoning, evidence, and uncertainty.
Repository status
The current repository includes:
the always-loaded workflow rules in
AGENTS.md;four parent skills and sixteen criterion skills under
skills/;four human-facing phase guides;
a project-level workflow tracker template.
The following surfaces are still planned:
project initialization and document templates;
companion subagent definitions;
continuity and guardrail hooks;
packaged installation and update commands.
The skills can be inspected and tested now, but the repository does not yet provide a complete installer. Projects using the workflow must currently place the skills and project files through their agent harness's normal configuration mechanism.
Repository structure
AGENTS.md
README.md
skills/
inquiry-analysis/
README.md
start-inquire-and-analyze/
explain-and-justify-the-need/
identify-and-prioritize-research/
analyze-prior-art/
develop-design-brief/
developing-ideas/
README.md
start-develop-ideas/
develop-design-specifications/
develop-feasible-ideas/
present-and-justify-the-chosen-design/
develop-planning-drawings-and-diagrams/
creating-solution/
README.md
start-create-solution/
construct-a-logical-plan/
demonstrate-technical-skills/
follow-the-plan-to-create-the-solution/
justify-changes-to-the-design/
evaluating/
README.md
start-evaluate/
design-testing-methods/
evaluate-against-the-design-specification/
explain-how-the-solution-could-be-improved/
explain-the-solutions-impact/
workflow-management/
workflow-tracker.mdThe phase folders organize the skills; they are not additional workflow phases.
workflow-management/ contains cross-phase project resources.
Sources and acknowledgments
Structured Workflow adapts the four criteria of the IB MYP Design Cycle for human-agent software work.
The criterion structure was checked against the IB MYP Design Subject Brief and the diagrams published by Design and Inquiry.
The workflow and skill design were informed by:
OpenAI's Product Design skill pack
The human-judgment and evidence-trace principles were also informed by Shaw and Nave's research on cognitive surrender.
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- Alicense-qualityDmaintenanceEnforces engineering governance for AI-driven software projects, ensuring state over prompt, freeze over generate, and audit over output through the 5S workflow.1MIT
- Flicense-qualityDmaintenanceEnforces a rigorous spec-driven engineering workflow for AI coding assistants, requiring requirements, design, and task planning before any code is written.
- Alicense-qualityCmaintenanceProtocol-enforced learning system combining memory-augmented reasoning with workflow automation to improve AI assistant reliability by ensuring they learn from past experiences before making code changes.8MIT
- AlicenseCqualityCmaintenanceEnforces code quality, security, and architectural integrity by auditing, refactoring, and documenting codebases within Cursor and Claude.10MIT
Related MCP Connectors
Adaptive plan/build/review cycles for AI coding assistants, persisted across sessions.
Verifies AI agent work end to end: real artifacts and outcomes checked, not self-reported success.
33 tools that make AI write, implement, and verify intent against explicit, testable constraints.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/kingdomseed/structured-workflow'
If you have feedback or need assistance with the MCP directory API, please join our Discord server