|
| 1 | +--- |
| 2 | +title: Customize tools |
| 3 | +description: |
| 4 | + Select which tools from an MCP server are available to your AI clients using |
| 5 | + the ToolHive UI. |
| 6 | +--- |
| 7 | + |
| 8 | +ToolHive lets you customize which tools from a running MCP server are available |
| 9 | +to your AI clients. This gives you fine-grained control over the capabilities |
| 10 | +exposed by each server, helping you create focused tool sets and reduce |
| 11 | +complexity for your AI workflows. |
| 12 | + |
| 13 | +## Why customize tools? |
| 14 | + |
| 15 | +Customizing tools helps you: |
| 16 | + |
| 17 | +- **Reduce complexity**: Hide tools that aren't relevant to your workflow, |
| 18 | + making it easier for AI clients to select the right tool |
| 19 | +- **Improve performance**: Fewer tools mean faster tool selection and reduced |
| 20 | + token usage in AI interactions |
| 21 | +- **Enhance security**: Limit exposure to only the tools you need, reducing the |
| 22 | + risk of unintended actions |
| 23 | +- **Create focused environments**: Tailor tool availability to specific tasks or |
| 24 | + projects |
| 25 | + |
| 26 | +:::info[Registry servers only] |
| 27 | + |
| 28 | +Tool customization is available only for MCP servers installed from the |
| 29 | +registry. Custom local and remote MCP servers do not support this feature. |
| 30 | + |
| 31 | +::: |
| 32 | + |
| 33 | +## Access tool customization |
| 34 | + |
| 35 | +To customize tools for a running MCP server: |
| 36 | + |
| 37 | +1. Open the **MCP Servers** page. |
| 38 | +1. Find the server you want to customize and click the menu (︙) on its card. |
| 39 | +1. Select **Customize Tools** from the dropdown menu. |
| 40 | + |
| 41 | +The customize tools page displays all available tools from the MCP server, each |
| 42 | +with a toggle switch and description. |
| 43 | + |
| 44 | +:::note[Server must be running] |
| 45 | + |
| 46 | +The MCP server must be running to view and customize tools. If the server is |
| 47 | +stopped when you access the customize tools page, you'll see a message prompting |
| 48 | +you to start the server first. |
| 49 | + |
| 50 | +::: |
| 51 | + |
| 52 | +## Enable or disable tools |
| 53 | + |
| 54 | +On the customize tools page: |
| 55 | + |
| 56 | +1. Review the list of available tools and their descriptions. |
| 57 | +1. Use the toggle switches to enable or disable individual tools: |
| 58 | + - **Green (enabled)**: The tool is available to AI clients |
| 59 | + - **Gray (disabled)**: The tool is hidden from AI clients |
| 60 | +1. Click **Apply** to save your changes. |
| 61 | + |
| 62 | +The ToolHive proxy filters out disabled tools when AI clients connect to the |
| 63 | +server. Clients only see and can only call the tools you've enabled. |
| 64 | + |
| 65 | +:::warning[At least one tool required] |
| 66 | + |
| 67 | +You must keep at least one tool enabled. ToolHive prevents you from disabling |
| 68 | +all tools on a server, as this would make the server non-functional. |
| 69 | + |
| 70 | +::: |
| 71 | + |
| 72 | +## How filtering works |
| 73 | + |
| 74 | +When you customize tools: |
| 75 | + |
| 76 | +1. You toggle tools on or off using the switches. |
| 77 | +1. After clicking **Apply**, your settings are saved to the server |
| 78 | + configuration. |
| 79 | +1. The ToolHive proxy intercepts tool discovery requests from AI clients. |
| 80 | +1. Only enabled tools appear in the client's tool list. |
| 81 | +1. If a client attempts to call a disabled tool directly, the proxy blocks the |
| 82 | + request. |
| 83 | + |
| 84 | +This filtering happens transparently without requiring any changes to your AI |
| 85 | +client configuration. Your clients automatically see the updated tool list the |
| 86 | +next time they connect. |
| 87 | + |
| 88 | +:::tip |
| 89 | + |
| 90 | +If you change tool settings while a client is connected, you may need to restart |
| 91 | +the client session to see the updated tool list. Some clients cache tool lists |
| 92 | +and don't automatically refresh. |
| 93 | + |
| 94 | +::: |
| 95 | + |
| 96 | +## Example workflows |
| 97 | + |
| 98 | +### Focus on specific GitHub operations |
| 99 | + |
| 100 | +If you're using the GitHub MCP server but only need pull request tools: |
| 101 | + |
| 102 | +1. Open **Customize Tools** for the GitHub server. |
| 103 | +1. Disable all tools except those related to pull requests (like |
| 104 | + `create_pull_request`, `get_pull_request`, `list_pull_requests`). |
| 105 | +1. Your AI clients now see only pull request tools, making it easier to work |
| 106 | + with GitHub PRs without distraction from issue, branch, or repository tools. |
| 107 | + |
| 108 | +### Create environment-specific tool sets |
| 109 | + |
| 110 | +If you're running the same MCP server in different groups for different |
| 111 | +environments: |
| 112 | + |
| 113 | +1. Copy the server to multiple groups (see |
| 114 | + [Organize servers into groups](./group-management.mdx)). |
| 115 | +1. Customize tools in each group to match the environment's needs. |
| 116 | +1. For example, enable only read-only tools in a production group, while |
| 117 | + allowing all tools in a development group. |
| 118 | + |
| 119 | +## Reset to defaults |
| 120 | + |
| 121 | +To restore all tools to their default enabled state: |
| 122 | + |
| 123 | +1. Open the customize tools page for the server. |
| 124 | +1. Manually enable all tools using the toggle switches. |
| 125 | +1. Click **Apply** to save your changes. |
| 126 | + |
| 127 | +There is no automatic "reset all" button, but you can quickly enable all tools |
| 128 | +by toggling each switch and clicking **Apply**. |
| 129 | + |
| 130 | +## Related information |
| 131 | + |
| 132 | +- [Run MCP servers](./run-mcp-servers.mdx) |
| 133 | +- [Organize servers into groups](./group-management.mdx) |
| 134 | +- [Client configuration](./client-configuration.mdx) |
0 commit comments