Case study
Agent workflow MCP server
A remote MCP server that gives AI coding agents one playbook, from onboarding a codebase to opening a pull request
- Client
- My product
- Role
- Built it
- Period
- 2026
Overview
A remote MCP server that turns how I work with AI coding agents into tools any MCP client can call. It prepares a codebase for agents, writes the guide they read first, and runs every ticket from plan to pull request the same way, on any project.
AI coding agents are only as good as what they know about a project and the checks around them. Without a shared playbook, every repository and every session starts from zero, and quality depends on who wrote the prompt.
What I did
- A remote MCP server on Cloudflare Workers that Claude Code and other MCP clients connect to once, behind an access token
- A readiness audit that checks a project for tests, CI, a staging branch and branch protection, and fixes the critical gaps before any agent work starts
- An onboarding workflow that writes the agent guide, the docs and a memory folder each project keeps in its own repository
- Ticket workflows: draft a ticket from a loose idea and check it against the current code, then run it from plan to a reviewed pull request
- A Next.js starter workflow and a design handoff workflow for Figma
- Templates that extend the same approach to marketing, design and product teams
- Hexrate runs on it: its agent guide, docs and memory came out of these workflows
