What is brand-loom?

brand-loom is a Python library of common marketing writing tasks, called skills, such as generating hooks, captions, hashtags, SEO outlines, FAQs, calls to action, bilingual translation, and LinkedIn posts. It works with any of four language model providers, OpenAI, Anthropic, Gemini, or a local model through Ollama, and you bring your own API key. It does not require a coding agent like Claude Code or Cursor, though it can also install into one. A quickstart with no API key at all uses a fake provider so you can try a skill immediately after pip install. With a real key, you set an environment variable for the provider and export the matching API key, then either call the Python functions directly, run a skill from the command line, or chain several skills together, such as generating a hook and then a caption from the same topic. Each skill can also take an optional brand_context dictionary describing tone, audience, and phrases to use or avoid, though the README notes this repository ships only a basic default, automatic brand voice extraction and multi-platform publishing live in a separate hosted product called Neoxra, built by the same team, which this open-source library feeds into. The repository includes a parity evaluation that runs every skill against every available provider and reports pass or fail in a table, currently showing all skills passing across the fake, OpenAI, Anthropic, and Gemini providers. Skills can also be run with an optional self-check flag that applies rule-based lint checks, such as banned-phrase flagging and length validation, which only flag issues rather than scoring or retrying automatically. The project can also be installed as an agent skill into Claude Code, Cursor, or other skills.sh-compatible hosts using a single command. It is released under the Apache-2.0 license.