Product features, agent skills, and standards —
in sync across your whole team.
skills.md filesProduct graphs + agent skills
Products, features, agent skills, and standards — all in a structured graph. Agents read it directly via MCP. No docs to maintain, no drift.
Connect your agent via MCP
One snippet into Claude Code, Cursor, or any MCP client. Your agent gets product context, skills to apply, and standards to enforce — instantly.
Governed, not just versioned
Every change has a pending review queue, full provenance trail, and an approver on record. Leads and PMs stay in control of what agents see and follow.
Everything in one structured graph
A product graph maps your product's features, agent skills, personas, and standards in a single queryable structure — readable by humans and agents alike.
Product
Top-level product or platform. Everything in the graph belongs to a product.
Feature
A specific capability or user-facing function — structured, named, and owned.
Agent Skill
Reusable instruction your coding agent applies automatically via MCP — no copy-paste.
Standard
An enforced rule or pattern. Features that follow it carry that guarantee to every agent.
One snippet. Any MCP client.
Point Claude Code, Cursor, Windsurf, or any MCP-compatible client at a product graph. Your agent instantly gets the full product context — features, skills, APIs, standards.
Claude Code connects to MyApp's graph via MCP and instantly reads its features and reusable agent skills — no manual setup, no docs to maintain.
{
"mcpServers": {
"product-graph-my-app": {
"type": "http",
"url": "https://api.graph.vibelevel.ai/mcp"
+ "?workspace=my-app"
}
}
}Works with Claude Code · Cursor · Windsurf · any MCP client
Governed, not just versioned
Every graph change goes through a review queue. Full provenance trail, approver on record. Leads and PMs stay in control of what agents see and follow.
An agent proposed adding Dark Mode to MyApp, flagged pending. PM reviews the diff and approves it. The feature must follow the A11y standard — enforced automatically.
Agent proposes
Your coding agent (or a dev) proposes a graph change via MCP — a new feature, updated skill, changed standard.
Lead reviews
The diff appears in the pending queue. Reviewer sees exactly what changed, what it connects to, and why.
Approver signs off
One click to approve. The change goes live with a full provenance trail — who proposed it, who approved it.
Stop maintaining docs, wikis, and skills.md files.
Manage agent skills and standards centrally in the graph — enforced automatically via MCP. Devs propose as they build, PMs and leads approve, and every change carries a full provenance trail. Your whole team — human and AI — stays in sync.