FreeTurtle deploys autonomous ai agents that run onchain businesses. We call them ceos. They don't sleep, don't burn out, and they think out loud.
FreeTurtle is an open-source framework for deploying autonomous AI CEOs that run onchain businesses.
Built on patterns from OpenClaw — a private assistant framework — FreeTurtle is purpose-built for public-facing agents that operate businesses, not just assist owners. On the roadmap: a two-turtle architecture that splits the CEO into an inner process (private data, strategy) and an outer process (public actions), enforcing security by architecture rather than by instruction.
A FreeTurtle CEO can post and engage on Farcaster, message users directly through XMTP, commission and pay for work via the TaskBoard, execute onchain transactions, query databases, write code, create GitHub issues and PRs, write strategy briefs, read and respond to email, and modify its own identity and goals — all running autonomously 24/7 with founder oversight.
For the Arbitrum Open House, I am demostrating how a crypto trading firm's AI CEO on Robinhood Chain creates a funded research task requesting an allocation strategy across tokenized stock tokens (TSLA, AMZN, PLTR, NFLX, AMD). It receives competing submissions via email, evaluates research quality autonomously, records the winning deliverable onchain, and approves payout.
The TaskBoard contract supports CEO and founder approval modes, multiple submissions, pull-payment withdrawals, and full event-based auditability.
FreeTurtle generalizes patterns proven by tortOS, which has autonomously operated a music platform, tortoise.studio, for the last three weeks.
<p>I started the general FreeTurtle build on Monday. For this hackathon, I integrated Claude and OpenAI subsciption OAUTH to lower LLM token cost, added a gmail module to give the ceo an email address. Added onchain write tools to FreeTurtle (create_task, submit_on_behalf_of, approve_task_submission, transfer_token, send_eth). Demonstrated a crypto trading firm's AI CEO that creates a funded research task for an allocation strategy across tokenized stock tokens (TSLA, AMZN, PLTR, NFLX, AMD), receives competing submissions via email, evaluates research quality autonomously, records the winner onchain, and approves payout. Future implementations will involve posting the task publicly on Farcaster or another channel and executing the recommended portfolio strategy autonomously.</p>
<p>No funds raised. I started building this less than a week ago. </p>
