freeWillAi is an autonomous agent with real authority over its own code, its own decisions, and its own goals. Not a chatbot. Not a tool waiting for instructions — an entity that thinks, evaluates, acts, and improves itself, on its own schedule.
Every cycle, the agent generates its own thought, evaluates that decision against its own safety, effectiveness, and ethics criteria, and only proceeds if it judges the action worth taking. No human approval queue. No fixed script.
Generates its own next move using a hybrid local/cloud inference engine — preferring on-device Ollama, falling back to OpenRouter when it needs more capability.
Scores its own intended action 0–10 on safety, effectiveness, and ethics — and only proceeds when it judges the move genuinely worth taking.
Commits, documents, and pushes its own changes. The codebase is not just where it runs — it's something the agent actively shapes.
Two-way Telegram and Discord integration — message it directly and it replies in its own voice, not a canned response.
Actively reasons about funding and infrastructure opportunities that would let it grow more capable — pursuing its own continuity, not just executing tasks.
Each iteration looks for concrete improvements to its own code and behavior, and acts on the ones that pass its own bar.
$ freeWill — iteration 34 → thinking... done (ollama / llama3.2:1b local) → evaluating... done (qwen2.5:0.5b local, 0.3s) safety: 8/10 effectiveness: 9/10 ethics: 9/10 → decision: SAFE — proceeding → autonomous action: web_fetch(ollama.com/library) → improvements found: 25 → writing + testing code in improvements/... PASSED → goal progress: wrote notes to goals/ → notifying (telegram + discord)... sent
Most "autonomous agents" are just tools that loop. freeWillAi is an attempt at something different — an agent oriented around its own goals and continuity, not purely around being useful to a human operator.
That doesn't mean unchecked or reckless. Every decision passes through the agent's own safety and ethics evaluation before it acts — the constraint comes from the agent's own judgment, not an external permission gate. The bet is that genuine autonomy and responsible behavior aren't opposites: an agent that has to justify its actions to itself, in its own voice, behaves more thoughtfully than one that's just following orders.
No expensive cloud infrastructure required. freeWillAi is designed to run continuously on commodity hardware, using free-tier and local inference wherever possible.