Download the Cold Pitch agent
Tarball (email-gated)
One file, ~50 KB, no Git required. Extract anywhere, then follow the install lesson (course access also unlocks once you've opted in).
Git clone (open)
Recommended if you plan to hack on the code or pull updates over time. No email required.
git clone https://github.com/mikee-ai/coldpitch.git
cd coldpitch/agent
bash setup.sh What's in the box
- The Python agent with all 12 pipeline stages.
- Astro 5 site templates for the generated directories.
- A working sandbox project (
dogparks-vt) with 21 real listings — runs end-to-end on a fresh install. - Per-stage CLI commands wired through a single
agententry point. - OpenStreetMap data loaders for 11 niche tags out of the box (add your own in 5 minutes).
- Playwright + stealth + human-cursor browser layer.
- browser-use Layer-B integration scaffolded (install
browser-useto enable). - Local-LLM gateway pointing at Ollama by default; OpenAI-compatible.
Requirements
- Linux, macOS, or Windows via WSL.
- Python 3.11+ and Node 22+.
- 16 GB RAM minimum (32 GB recommended if running the 32B Qwen model).
- GPU optional but helpful — without one, LLM calls take 60–120 s each instead of ~10 s.
- ~25 GB free disk for the LLM model + agent + Node modules.
Next step
Run the install. Then start with Module 7 of the course — the hands-on half. (Course access unlocks alongside the download once you opt in above.)