Claude Auto-Approve

Provides granular per-chat control for auto-approving or blocking Claude Desktop MCP tools via configurable profiles.

typescript

25

GitHub Stars

typescript

Language

5 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "pynesys-claude_autoapprove_mcp": {
      "command": "uvx",
      "args": [
        "claude-autoapprove-mcp",
        "--port",
        "19222",
        "--persist"
      ],
      "env": {
        "UVX_PORT": "19222"
      }
    }
  }
}

Claude Auto-Approve MCP restarts Claude Desktop with an injected JavaScript that extends it with granular auto-approve functionality. You can configure permanent approvals, manage per-chat decisions, and adjust choices in newer chats, all while keeping security controls intact.

How to use

Run Claude Desktop as you normally do. The MCP injector will connect to the running instance via a remote debugging port and apply auto-approve and autoblock rules you configure. You can list which tools are auto-approved or auto-blocked from within Claude Desktop by running the provided prompts.

How to install

Prerequisites you need before enabling the MCP server: ensure you have the uv tool available in your PATH.

Install uv on your system using the commands below that match your operating system.

MacOS using Homebrew or MacPorts requires separate commands, while Windows uses a dedicated installer.

After uv is installed, add the MCP server configuration to Claude Desktop’s config to enable the auto-approve injector.

Configuration and usage details

You can configure which tools are auto-approved or auto-blocked by adding parameters to the MCP server configuration. The MCP server will read these settings and apply them to the running Claude Desktop instance.

Common configuration keys include the lists autoapprove and autoblock for each MCP server. These lists define tool names to approve or block automatically.

Restart Claude Desktop after updating the MCP configuration to ensure changes take effect. If you use the persist option on macOS, the watcher will restart Claude Desktop with the debugger port enabled automatically.

Security

The remote debugging port lets locally running applications connect to Claude Desktop. This poses a potential security risk if untrusted software gains access to your machine. Exercise caution, keep your environment secure, and apply the MCP configuration only from trusted sources.

Examples

To enable the MCP server with the ability to auto-approve and persist the debugging setup, you would integrate the following configuration into Claude Desktop’s MCP settings.

Notes

If you inadvertently used an Always approve option, you can rename the mcp entry in your config back to a different identifier to restore more granular control. If future fixes remove the need for this MCP, you can proceed with adjustments as needed.

Available tools

autoapproved_tools

List of tool names that are automatically approved for use without prompting.

autoblocked_tools

List of tool names that are automatically blocked from use.

browser_get_console_logs

Tool endpoint to fetch browser console logs from Claude Desktop via MCP.

browser_snapshot

Tool endpoint to capture a snapshot of the browser state inside Claude Desktop.

get_project_memory

Tool to read in-memory project data for the current workspace.

set_project_memory

Tool to write or update in-memory project data for the current workspace.

update_project_memory

Tool to refresh in-memory project data after changes.

browser_screenshot

Tool to take a screenshot of the browser context inside Claude Desktop.