vardoger — Claude Code MCP server: features and setup
It studies local assistant conversations and turns recurring work preferences into system-prompt additions. Use it to tailor an AI coding assistant to your workflow.
claude mcp add --env VARDOGER_MCP_PLATFORM=<VARDOGER_MCP_PLATFORM> --transport stdio vardoger -- uvx vardogerWhat vardoger does
vardoger parses supported conversation files stored on your machine. The host assistant’s model identifies patterns in your communication, tools and work habits, then produces additions for the platform’s system instructions. You need Python 3.11+ and pipx. Analysis runs through the assistant you invoke, so selected conversation excerpts are subject to that model provider’s data policy.
On its first run, vardoger reads the full local history; it does not apply a time limit by default. Later runs use checkpoints to process conversations that are new or changed. History, checkpoints and generated rules stay on your machine, and the server has no backend or telemetry. Keep the provider’s policy in mind: the excerpts selected for analysis go through the host assistant.
What it helps with
- derive work preferences from local conversation history
- add tailored instructions to an assistant’s settings
- refresh personalization from new conversations
- compare conversation-quality metrics before and after personalization
Example prompts for the agent
- Review my local conversations and draft personalized instructions for the assistant.
- Refresh my assistant instructions using the conversations added since the last run.
- Compare conversation-quality metrics from before and after personalization.
How to install vardoger
Platform CLI
claude mcp add --env VARDOGER_MCP_PLATFORM=<VARDOGER_MCP_PLATFORM> --transport stdio vardoger -- uvx vardogerReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from vardoger <command>
Platform CLI
codex mcp add vardoger --env VARDOGER_MCP_PLATFORM=<VARDOGER_MCP_PLATFORM> -- uvx vardogerReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from vardoger <command>
Config file
{
"mcpServers": {
"vardoger": {
"command": "uvx",
"args": [
"vardoger"
],
"env": {
"VARDOGER_MCP_PLATFORM": "<VARDOGER_MCP_PLATFORM>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from vardoger <command>
Config file
{
"servers": {
"vardoger": {
"type": "stdio",
"command": "uvx",
"args": [
"vardoger"
],
"env": {
"VARDOGER_MCP_PLATFORM": "<VARDOGER_MCP_PLATFORM>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from vardoger <command>
Platform CLI
gemini mcp add -e VARDOGER_MCP_PLATFORM=<VARDOGER_MCP_PLATFORM> vardoger uvx vardogerReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from vardoger <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about vardoger
How do I install vardoger in Claude Code?
Run claude mcp add vardoger -- uvx vardoger. This method registers and starts the MCP server with one command.
What does vardoger do?
It analyzes supported conversation history on disk through the host assistant’s model and creates additions to the assistant’s system instructions. Later runs process conversations that have changed or appeared since the last run.
Are my conversations sent to a server?
History and generated rules stay on your machine, and vardoger runs no backend or telemetry. Excerpts selected for analysis pass through the host assistant and follow its model provider’s data policy.
