Bring the agent you already use.
Holbric is a standard stdio MCP server. These hosts each have a one-line connect command, and any other host can use the printed config.
Claude Code
holbric connect claude
Runs claude mcp add --scope user holbric when the claude CLI is on your PATH. Otherwise it adds Holbric to ~/.claude.json. Start a new session and run /mcp.
Cursor
holbric connect cursor
Writes .cursor/mcp.json in the current project. Add --global to write ~/.cursor/mcp.json instead. Then turn on holbric under Settings → MCP.
Codex
holbric connect codex
Adds an [mcp_servers.holbric] block to ~/.codex/config.toml. Start a new session and run /mcp.
Gemini CLI
holbric connect gemini
Adds Holbric to mcpServers in ~/.gemini/settings.json. Restart Gemini CLI and run /mcp.
Windsurf
holbric connect windsurf
Adds Holbric to ~/.codeium/windsurf/mcp_config.json. Then refresh the MCP servers in Cascade.
VS Code
holbric connect vscode
Writes .vscode/mcp.json in the current project. Open Chat in agent mode and start holbric.
Other MCP hosts
holbric connect cursor --print
Prints the JSON and writes nothing. Any host that can run a local stdio command can use it as is.
What every host gets
The same 19 tools, the same rule of a summary before any signature, and the same local wallet. Holbric's server instructions tell the host to show you each summary word for word and to wait for a clear yes before calling execute_prepared.
execute_prepared out of it. That way the final yes stays with you.Using your own agent
Writing your own agent or automation? Start holbric mcp as a subprocess and talk MCP over stdio. Every write still needs a prepare call followed by execute_prepared with confirm: true. Use a separate wallet with a daily limit (holbric limit) for anything that runs without you watching.