How a system works
Architecture, logic, and how the pieces connect.
Connect Virgo to any MCP client. Your agents pull the decisions, conventions, and history behind your code, then save what they learn for the next session.
Your agents talk to the Virgo MCP server. It reads the reasoning graph behind your code and serves the right context back to whichever agent is asking. It also saves what agents learn back to the graph.
Virgo ships with its tools and the instructions for using them. As soon as an agent connects, it knows what Virgo offers and when to reach for it.
There is no skills file to write, no rules to maintain, and nothing to wire up. The configuration travels with the endpoint.
Architecture, logic, and how the pieces connect.
Decisions, incidents, and fixes over time.
Patterns recalled before code is written.
Written back to the graph for every agent.
MCP puts the reasoning graph where your agents do the work.
New engineers and new agents get the reasoning behind the code from day one. Weeks of asking around turn into targeted follow-up.
Follow a feature across tickets, PRs, and threads without stitching the story together by hand.
Agents write to your conventions and prior decisions instead of falling back on generic defaults.
Every agent reads and writes one shared reasoning graph. Learnings accumulate across sessions.
What one agent discovers is ready for the next, across tools, repos, and sessions.
Decisions made in chat and review reach your agents as they happen, before the next reindex.
Use the same endpoint everywhere. Add it with a config file or one command.
Add the endpoint from your project root with one command.
claude mcp add --transport http virgo https://org.mcp.virgo.fyi/
Virgo’s tools appear the next time you run Claude Code. Check /mcp.
Add the server to your Codex config file.
[mcp_servers.virgo] url = "https://org.mcp.virgo.fyi/"
Restart Codex and the Virgo tools are available to every session.
Add the server to your project MCP config.
{
"mcpServers": {
"virgo": {
"url": "https://org.mcp.virgo.fyi/"
}
}
}Cursor picks it up on save. Virgo appears under Settings, MCP.
Use the same endpoint everywhere. Add it with a config file or one command.
Add the endpoint from your project root with one command.
claude mcp add --transport http virgo https://org.mcp.virgo.fyi/
Virgo’s tools appear the next time you run Claude Code. Check /mcp.
Common questions about configuring and using MCPs.
The Model Context Protocol is an open standard that lets AI agents call external tools. Virgo runs an MCP server. Any compatible client uses it without a custom integration.
Any MCP client. We document Claude Code, Codex, and Cursor, and the same endpoint works in Copilot, Windsurf, and anything else that speaks MCP.
No, Virgo also exposes the same tools through standard REST APIs.
No. Virgo sends its tools and the instructions for using them when an agent connects. The agent knows what to reach for without any local setup.
They can pull context, trace causality, recall team conventions, and save new learnings back to the graph for every other agent and session.
Virgo serves the reasoning graph your team has already connected and runs inside your workspace boundary. See the security page for how data is handled.
About a minute. Copy your workspace endpoint. Add it to your client’s MCP config or run one command. The tools appear automatically.