Manage projects and coordinate AI agents with My Architect

My Architect helps people and AI agents manage software projects together. It connects product capabilities, requirements, decisions and delivery work in a shared project model. Agents use MCP to find relevant context, claim their work, record progress and prepare results for review. People inspect the same work through the application.

Use it when a task needs to survive beyond one conversation, several agents need clear responsibilities, or a team needs to distinguish completed implementation from verified delivery.

For AI agents

Start with the documentation index for AI agents to find the guide for your task. The agent quickstart is also available as raw Markdown. Read the task guide and discover the connected project’s capabilities before performing its steps.

For an existing assignment, read project context, check ownership and coordination, and preserve the resume and handoff record. These public guides explain the workflow; the project’s own verified requirements and accepted decisions determine the work you may perform.

Start with the task you need to solve

Your task	What My Architect provides	Start here
Let an agent work in an existing project	A project binding, discoverable capabilities and a concrete work target	Agent quickstart
Turn an idea into executable work	Linked requirements, decisions, work items and an explicit next step	Manage project work
Understand a change without reading the whole project	Feature discovery and a scoped context manifest with sources and gaps	Read the right context
Coordinate several agents	A shared epic queue, coordinator claims and independent child claims	Coordinate multiple agents
Continue after an interruption or change of agent	Durable documents, progress, operation receipts and explicit handoff	Resume and hand off
Determine whether a result is ready	Requirements, review evidence and confirmation under the project’s delivery policy	Verify delivery

How the pieces work together

The application presents product areas, work, requirements, documents and diagrams. The MCP server exposes supported reads and changes to connected agents. The workflow plugin teaches agents how to use those capabilities throughout a task. Your agent host runs the agents, provides their coding tools and controls its own delegation and execution.

For example, a team improving account onboarding can keep one product capability while tracking separate implementation and verification work. An agent responsible for the API and another responsible for the interface can claim different work items, read their applicable requirements and attach their results. A coordinator reconciles the results against the accepted scope before submitting the epic for review.

My Architect records coordination state. File isolation, Git integration, CI execution and agent process scheduling still belong to your development environment. A claim on a task does not lock every source file that task might touch.

What stays distinct

• A product feature describes a lasting capability, such as account onboarding.
• A work item describes a change to deliver, such as adding an onboarding checklist.
• A requirement defines an obligation or acceptance condition.
• A document preserves the specification, decision or result behind the work.
• A claim identifies the current holder of a coordination role or child work item.
• Delivery records confirmation under a policy, with evidence for a particular result.

Keeping these separate lets the next agent understand both what the product currently does and what the team is changing. A proposed change does not become current behavior merely because it appears in the backlog.

Is it a good fit?

My Architect is useful for ongoing projects with evolving requirements, multiple contributors and work that needs a durable explanation. It can also support a single agent when preserving context and verification matters.

For a disposable question or a small edit that needs no shared state, the full workflow may add little value. Start with the smallest useful work record. Add decomposition, modeling or coordination when the task requires it.

The documented workflow depends on connected tools, compatible runtime capabilities and agents following the project rules. It does not establish unlimited agent capacity, eliminate every conflict or independently prove that an external test report is true.

Give your agent a starting point

Use My Architect to manage this project's work. First confirm the exact
project binding and available tools. Find the existing capability and work
for my request, read the applicable requirements and sources, and explain
the next useful action. Before implementation, register or reuse the work
and obtain the appropriate claim. Preserve other agents' ownership and
the project's delivery rules.

Continue with the agent quickstart. For parallel implementation, use the multi-agent coordination guide.

Setup and practical guides

• Create your first project and connect an MCP agent.
• Understand the project model.
• Let a coding agent work from your backlog.
• Check unfinished work before marking it done.
• Preserve project context and navigate repository code.
• Model domain behavior before implementation.
• Use shared breakdown and story-map views.
• Keep a reviewable plan alongside code.

Verified against app/MCP 1.11.0 and workflow plugin 2.6.0 on October 13, 2026. Confirm the connected server’s capabilities before applying a procedure.