Skip to main content
Contact PuglieseWeb on WhatsApp or Call

Private preview In daily use at PuglieseWeb · not yet open to the public

SystemDox MCP

What your AI reads before it writes a line of code.

Your tech stack, checks, rules, specs and documentation, served to every coding agent over the Model Context Protocol at write time — and the decisions the agent makes written back into the library.

Request early access All modules
Standards: Tech Stack, Checks, Model Settings, Rules and Prompt Templates 
{
  "mcpServers": {
    "systemdox": {
      "type": "http",
      "url": "https://<your-workspace>.mcp.systemdox.com/mcp",
      "headers": {
        "Authorization": "Bearer ${SYSTEMDOX_API_KEY}"
      }
    }
  }
}
get_context report_decision update_spec npx @systemdox/mcp-server setup
The configuration block for Claude Code, Cursor or any MCP client. The key lives in an environment variable and is scoped to your workspace.

  1. Standards declared once

    Tech Stack, Checks, Model Settings, Rules and Prompt Templates. Declare them in SystemDox and every assistant reads the same set.

  2. A hosted endpoint per workspace

    Point Claude Code, Codex, Cursor or any MCP client at your workspace's endpoint. Nothing to install, one configuration block, a key scoped to your tenant.

  3. Ranked, within a token budget

    get_context ranks the checks, specs and docs relevant to the task and fetches their bodies within a token budget, so the agent starts from your architecture instead of guessing it.

  4. Both directions

    Agents call report_decision and update_spec as they work. Decision records exist as a by-product of shipping, not as archaeology afterwards.

  5. Every agent's configuration, generated

    Model Settings emits the configuration for each tool you use — for Claude Code that is CLAUDE.md, settings and permissions, hooks, skills and sub-agents — from the standards you authored once.

  6. Checks become fitness tests

    Publish your checks to GitHub as architecture fitness tests that run on every pull request. Failures come back through Insights as the next standard.

SystemDox MCP

Declare what your AI must follow

Tech Stack: three languages and fourteen frameworks declared with versions
Checks: 27 checks waiting to be published to GitHub as architecture fitness tests
Model Settings: Claude Code — CLAUDE.md, settings and permissions, hooks, skills and sub-agents
Rules: six deployment rules and two security rules
Prompt Templates: adding a reusable template with a name, description and prompt
Insights: the quality gates and the Report an Issue box that turns a mistake into a standard

At PuglieseWeb

Claude Code works on all of our repositories with SystemDox MCP connected. Our stack, six deployment rules, two security rules and the Claude Code configuration you see here are the ones it reads.

Every screenshot on this site is from our own workspace as it is today, not a mock-up. What you see is what runs.

Want this for your team?

SystemDox is in private preview. Request early access and we will write to you personally when a place opens.

Request early access
Next module SystemDox Knowledge