# Custom Node Manager

The Custom Node Manager installs, updates, and removes community node packs directly from the InvokeAI UI — no manual file copying, no restart required.

## Opening the Custom Node Manager

Click the **Nodes** tab (circuit icon) in the left sidebar, between **Models** and **Queue**.

The page is split into two panels:

- **Left:** the list of installed node packs, with each pack’s node count, type badges, and on-disk path.
- **Right:** the install UI, with tabs for **Git Repository URL** and **Scan Folder**, plus an install log at the bottom.

## Installing a node pack

1. On the right panel, choose the **Git Repository URL** tab.
2. Paste the Git URL of the pack, e.g. `https://github.com/user/my-node-pack.git`.
3. Click **Install**.

What happens during install:

- The repo is cloned into your `nodes` directory.
- The nodes are loaded into the running InvokeAI process immediately — **no restart needed**.
- Any workflow `.json` files found in the repo are imported into your workflow library and tagged with `node-pack:<pack-name>` so you can filter for them.
- The install log at the bottom of the panel shows the result for each step.

### Python dependencies

The Custom Node Manager **does not** automatically run `pip install` for a pack’s `requirements.txt` or `pyproject.toml`. Auto-installing into the running InvokeAI environment risks pulling in incompatible package versions and breaking the application.

If a pack ships extra dependencies, you’ll see a warning toast after installation. Install them yourself — typically `pip install -r requirements.txt` from inside an activated InvokeAI environment, but check the pack’s README first. After installing, click **Reload** so the new dependencies take effect.

## Managing installed packs

Each entry in the left panel has actions for managing the pack:

- **Reload** — re-scans the `nodes` directory. Use this after manually adding a pack via `git clone`, or after installing extra Python dependencies.
- **Uninstall** — removes the pack from disk, unregisters its nodes from the running process, and removes any workflows that were imported from the pack. No restart needed.

## Scan Folder tab

The **Scan Folder** tab shows the path of your `nodes` directory. Anything placed there manually (for example, by `git clone`-ing a pack directly) is detected automatically at startup. Use **Reload** to pick up packs added at runtime.

## Troubleshooting

### Install fails

- Confirm the Git URL is correct and reachable.
- The repo must contain an `__init__.py` at its root.
- Read the install log — it surfaces the underlying error.

### Nodes don’t appear after install

- Click **Reload**.
- Check that the pack’s `__init__.py` imports the node classes.
- Check the server console for import errors.

### Workflows show errors after uninstalling

User-created workflows that reference nodes from an uninstalled pack will show errors for the missing node types. Either reinstall the pack or remove the affected nodes from the workflow.

## Authoring a node pack

If you want to publish your own pack so it can be installed by URL, see the [Creating a Node Pack](/content/development/guides/creating-nodes/index.html) developer guide for the required repository layout, `__init__.py` requirements, and conventions for shipping workflows alongside your nodes.