Architect operates two US brokerages, an international perpetuals exchange, and soon a US commodities futures/options exchange with an engineering team of 12. We do this with a tailored fleet of agents we built from scratch. Here's a partial list:
• An agent with permissioned access to our exchange databases that can be asked from Slack or Claude to generate charts, tables, and other business-intelligence reports. Built heavily on ClickHouse and Matplotlib.
• An agent that reads the past week of X and media coverage, our internal Slack, and our exchange-transactions DB to write a weekly internal newsletter on the exchange landscape and what we should focus on next.
• An agent that uses the X API and HeyGen's HyperFrames to post a daily summary of GPU compute prices with an accompanying motion-graphic video. Compute is the core commodity of the AI economy we're building an exchange for.
• An agent that acts as the team's project manager, using Linear to summarize key milestones, track development velocity, and triage lagging or deprecated tickets.
• An agent that scans for new waitlist entries and posts a breakdown to Slack, grouped by firm and individual, with their stated interest.
• An agent that scans our Rust and TypeScript dependencies for vulnerabilities and automatically opens PRs to upgrade or remove affected packages, tuned to our stack rather than off-the-shelf tooling.
• An agent that shepherds GitHub pull requests that have gone more than a day with unaddressed comments.
We're scaling this fleet to create the first US derivatives exchange purpose-built for AI and US reindustrialization. If there are agent functions you've found useful in this area, let us know here or in Architect’s compute markets Telegram.