A Workspace Tool gives an Open WebUI chat model a server-side Python action it can call during a conversation. Use one when a model needs a controlled capability outside normal text generation, such as formatting domain data or returning a result from an approved internal service.
Workspace Tools live under Workspace → Tools, separate from Functions and external MCP or OpenAPI tool servers. An administrator or a user with permission to create tools can install a Python file containing a metadata docstring and a Tools class. The connected chat model must support native tool calling.
Tool source executes on the Open WebUI backend, so installation carries the same trust requirements as deploying application code. Untrusted source or dependencies can access server resources with the backend's privileges. The checksum tool hashes a supplied phrase without reading files, using credentials, or contacting another service.
Steps to install an Open WebUI Workspace Tool:
- Open Workspace from the sidebar.

- Select the Tools tab.

- Click Create to open the tool editor.

- Enter Tool Install Smoke in Tool Name.

- Set Tool ID to tool_install_smoke.
The ID is a lowercase Python identifier containing letters and underscores.
- Enter Returns the SHA-256 digest of the supplied phrase. in Tool Description.

- Replace the starter code with the metadata docstring.
""" title: Tool Install Smoke author: Open WebUI admin version: 0.1 description: Returns the SHA-256 digest of the supplied phrase. """
- Import sha256 from Python’s standard library below the docstring.
from hashlib import sha256
- Add the Tools class below the import.
class Tools: pass
The class groups the functions Open WebUI exposes to the model. The temporary pass keeps the empty class valid Python.
- Replace pass with the indented tool method.
async def sha256_phrase(self, phrase: str) -> str: """Return the SHA-256 hex digest of phrase encoded as UTF-8.""" return sha256(phrase.encode("utf-8")).hexdigest()
The type hint exposes phrase as a string argument. The method hashes the supplied UTF-8 text. Async methods are recommended for tool compatibility.
Saving loads Python code on the server. Only trusted code belongs here; third-party tools require a source and dependency review before installation.
- Click Save & Create.

- Click Confirm to install the reviewed tool.
The saved tool appears in the Tools list. Creating it does not automatically select it in a chat.
- Open New Chat from the sidebar.

- Open the model picker in the message composer.

- Select a model that supports native tool calling.
The selected model here is company-chat:latest. The connected provider determines the available model names.
- Open Integrations using the four-diamond icon beside the + button.

- Select Tools in the Integrations menu.

- Enable Tool Install Smoke for the chat.

- Enter a request with a distinctive phrase in the message box.
Call sha256_phrase with phrase blue-sky-router-417. Report the returned digest.
- Send the request.

- Expand View Result from sha256_phrase in the reply.

- Confirm that the reply reports the digest shown in the tool’s OUTPUT.
The tool’s INPUT must contain blue-sky-router-417. Its OUTPUT is a 64-character digest, and the reply must report that same value. A sentence claiming the tool ran without a tool-call entry is insufficient.