Bonfire
Your Claude Code, opinionated. Pip-install a runtime and a Claude Code skill; run /bonfire scan to wire a nine-role cadre around your project.
Open-source tools — Bonfire and CandyFactory OS — and custom AI systems for the businesses that run on them. The discipline is the method, not the headcount.
Your Claude Code, opinionated. Pip-install a runtime and a Claude Code skill; run /bonfire scan to wire a nine-role cadre around your project.
A trademark-clean, reproducibly-built Linux distribution tuned to be the substrate our agents run on — branding now, kernel-level levers next.
An ops-drift detector for the whole surface: it watches the project the way a project manager would and flags what slipped. A Bonfire-substrate plugin.
A build-in-public engine for solo founders: it mines your build log for moments and drafts platform-shaped posts under your own voice. A Bonfire-substrate plugin.
A review-acceleration surface: review the running program at claim and architecture altitude — flow over diff, where glow means heat means risk. A Bonfire-substrate plugin.
An engagement is the method applied to one operation: an AI system designed around how your business actually works, built under structural TDD, and run as a managed service. Not a template you configure — a system tuned to you, owned by you, kept working by us.
Same claim, three altitudes: the open source proves the method in public, the shop applies it to a business outcome, and a bespoke engagement sells it tuned to one operation.
Your Claude Code, opinionated. Pip-install a runtime and a Claude Code skill, then run /bonfire scan to wire a nine-role agent cadre around your project — structural TDD and audit-driven release gates, the discipline of a real engineering team applied to AI.
It is the open-source proof of the method: the same cadre we run for clients, packaged for anyone to install with pip install bonfire-ai.
A trademark-clean, reproducibly-built Linux distribution tuned to be the substrate our agents run on. Branding and a Cotton Candy Parlor theme now; kernel-level levers — tuned config, hugepages, CPU pinning, io_uring — next.
Built by a headless reproducible factory line, not a GUI, so the whole image compiles the way the rest of the work does.
A cross-surface ops-drift detector. It watches the project the way a project manager would — comparing the documented plan against what actually happened — and flags what slipped, on a weekly and on-demand cadence.
A plugin on the Bonfire substrate: the same audit-driven discipline, pointed at the operational fabric instead of the code.
A self-hostable build-in-public engine for solo founders. It mines your build log for moments, drafts platform-shaped posts for LinkedIn, X, Instagram and TikTok, and audits each one against your own voice and the platform rules before you queue it.
A plugin on the Bonfire substrate: your story, drafted with the same discipline as the code.
A review-acceleration surface. Review the running program at claim and architecture altitude — flow over diff, where the view is the compilation and glow means heat means risk — so an untreated exception lights up instead of hiding in a thousand-line patch.
A plugin on the Bonfire substrate: faster, higher-altitude review of what the program actually does.
An engagement is the method applied to one operation: an AI system designed around how your business actually works, built under structural TDD, and run as a managed service. Not a template you configure — a system tuned to you, owned by you, kept working by us.
Same claim, three altitudes: the open source proves the method in public, the shop applies it to a business outcome, and a bespoke engagement sells it tuned to one operation.
Your intent, in plain words. One person, steering.
Turns your intent into precise missions — then conducts the cadre. The orchestration is the part that isn't “just agents.”
↓ conducts the cadreReal, tested software — shipped only when you say so. Then the line loops.