How to use Fabric: install, run patterns and set up providers

What Fabric is and who it is for

Fabric organizes prompts by real-world task so you can create, collect and reuse them in one place. The project describes its goal as fitting AI into daily work instead of switching between separate chat apps.

The README centers on the command line, so Fabric suits people comfortable with a terminal and configuration files. It is less suited to anyone who wants a point-and-click chat app.

• Good fit: people who want to run AI prompts on files or piped text
• Good fit: people who want to collect and reuse their own prompts
• Good fit: developers who want a REST API server built on the same patterns
• Not a fit: anyone without access to an AI provider account, since Fabric connects to model backends rather than supplying its own

Install Fabric from the repository

The README lists a package-manager install method and does not list any prerequisites. Its install command shown in the source clones the repository from GitHub. Check the README for the full set of install options for your system before you start.

After cloning, follow the install instructions in the README and the docs folder. Some older videos were recorded when Fabric was Python-based, so rely on the current README instructions.

Run a pattern on your own text

Patterns are the core unit of Fabric. Pick the pattern that matches your task, then send it your input from the command line. The README's screenshot shows a summarize pattern being run, which is a good first example.

Fabric also accepts piped text, so you can feed it output from other command-line tools. The README does not list every flag, so check its usage section and documentation for exact syntax.

• Choose a pattern that matches the task, such as summarizing
• Send your text as input from the command line or through a pipe
• Review the output and adjust the pattern or input if needed

Create and organize your own patterns

Fabric lets you collect and organize your own prompts for recurring tasks. The README includes a Custom Patterns section that covers how to write them.

Once a prompt works well for a task, keeping it as a pattern means you can reuse it instead of rewriting it each time.

Connect an AI provider

Fabric connects to AI model backends. You need your own account or access with a provider, because Fabric does not supply models itself. The README's recent updates list many supported providers, and the project's changelog shows that the list changes over time.

Use the provider setup described in the README and docs. The project also notes that a REST API server can expose the same patterns to other programs.

• Confirm you have access to at least one provider before installing
• Check the README for the current list of supported providers and setup steps
• Use the REST API server if you want other tools to call your patterns

Where to find more detail

The README points to the docs folder for a deep dive into internals and to DeepWiki for an AI-generated guide to the project. The README also links to videos, though some were recorded for an earlier Python version of Fabric.

Because the README is the main reference, new users should expect to read the docs and configuration notes themselves rather than rely on a step-by-step beginner guide.

• Docs folder: https://github.com/danielmiessler/Fabric/tree/main/docs
• DeepWiki: https://deepwiki.com/danielmiessler/Fabric