AI coding agents

Product context for Claude Code, Cursor and every coding agent

A CLAUDE.md holds the repo's rules; it cannot hold the company. Epic Context gives Claude Code and Cursor the product's design system, voice, users, features and decisions through one MCP connection, the same brain the rest of the team works from.

The gap

A CLAUDE.md holds the repo's rules. It cannot hold the company.

CLAUDE.md, AGENTS.md and Cursor rules carry the code's conventions: how to build, test and name things. The brand voice, the personas, the design system and the decisions behind a feature live somewhere else, and they drift.

So every session starts with pasting the brand guidelines again, and the agent still builds a generic screen.

Before every task

The agent reads a briefing, then it builds.

The agent asks for a briefing for its task and reads what that kind of task must see, each block with its version. A write gate checks that it read before it wrote.

  1. The ask

    Build the settings page.

  2. It reads first

    • BuildStack profile
    • DesignComponent spec
    • DesignColour system
    • BrandTone of voice
    • ProductPage pattern
  3. It works as

    The Design expert

    • The write gate checks that it read first
    • The page's block in the brain lands in the same change
  4. What comes back

    A settings page in the product's own system

    • Built from the components and tokens the specs name
    • Copy in the tone of voice
    • The page recorded in the brain beside the code
Our own useHow Epic Context itself is built: a page task in this product starts with this briefing.

Context as you build

Code and context ship together.

When the agent builds UI, the page's block lands in the brain in the same change: what the page is, why it exists and who it is for. The next session, and the next colleague, see what was built and why.

The plugin

The plugin brings the skills and the hooks.

The Epic Context plugin adds skills and hooks to Claude Code: the experts' methods, a line at the start of every session that anchors the agent to its project, and a reminder before it stops. A repo is bound to its project by a committed .epicontext file.

The plugin comes from our public marketplace on GitHub: anyone can install it, and it works once you are invited to Epic Context.

For the product owner

Stop re-briefing every session.

The product owner keeps the brand, the personas and the decisions right once, in the brain, and every coding agent reads them. Nobody pastes the same guidelines into the next session.

Stakeholders walk the review build with the Epic Context widget open and comment on any page; the comment lands against that page's block, with a reason, in the changes feed.

Questions

How do I give Claude Code my product context?

Connect Claude Code to Epic Context over MCP with your company account, then let it ask for a briefing before each task. It reads the design system, voice, users and decisions it needs, and writes back with a reason. The plugin adds the skills and hooks.

Does it replace CLAUDE.md?

No. Keep CLAUDE.md or AGENTS.md for the repo's own rules. Epic Context holds what a repo file cannot: the company's brand, users, design system and decisions, shared across every repo and every tool.

Does it work with Cursor?

Yes. Cursor connects over the same MCP connector and reads the same brain as Claude Code and Claude.

Can the agent change the context?

Yes, with a reason. Every write is checked against the schema, shows up in the changes feed with its source, and can be undone in one click. A locked block refuses every writer.

How do product managers use it with AI?

They keep the product's truth in one place: personas, journeys, features, decisions and the brand. Their own AI sessions, the developers' coding agents and the stakeholders' comments all work from it, so nobody has to re-brief each tool.

Start with one real project

Epic Stories installs Epic Context with your team, on your own work, and then steps back. Tell us where you would start.