# Bautista Filippini > I agentize companies. Co-founder of Prophet. I turn the work a company already does into tools its AI agents can use. > Buenos Aires, Argentina (UTC−3). Open to remote. Canonical URL: HTML version: ## What I do I build internal software, shared memory for teams, AI agents, tools for AI agents, decision dashboards and MCP servers. ## Prophet: we agentize companies I am a co-founder of [Prophet](https://prophet.lat), a two-person company in Buenos Aires. Prophet agentizes companies. We take the processes and systems a company already runs and add a layer on top that turns each task into a tool an AI agent can use. Processes become skills, and the skills live in the company's own plugin or MCP server, so it works from Claude, ChatGPT or any other agent. We start with small processes: first an agent does the task while a person approves it, then it runs on its own. - **Your tasks become tools.** Each process is packed as a skill an agent can run. - **Nothing gets replaced.** The layer sits on top of the systems you already use. - **Small first, then on its own.** A person approves at the start; then it runs alone. ## In production — seven systems people use every day ### One Brain Shared memory for teams working with Claude Code — my own product. Each session writes down what it closed, so the next one starts knowing what the rest of the team did. Multi-tenant, Google login, MCP plugin. Real teams are using it. - Stack: Next.js, Supabase, MCP, PostgreSQL - Status: 1,470 tests green - URL: ### Plata en Mano The whole credit lifecycle for a finance company, from origination to arrears. Origination, instalments, collections and arrears, plus an investor portal where each investor sees their own capital, returns and overdue amounts. A companion app fills their pledge paperwork automatically, to the millimetre, from the same data. - Stack: Next.js, PostgreSQL, row-level security, PDF engine - Status: 90/90 tests green, reconciled to their books ### Real estate MCP An agency asks its own Claude about the market, and the answers come from real listings. The server tells them what a property has and what comparable ones are asking — and says "I don't know" when the data doesn't back an answer. Hosted, so each agency connects it to its own Claude. - Stack: MCP server, Python, scraping pipeline - Status: 10,351 properties indexed - URL: ### Purchase app A wool buying team's spreadsheet, replaced by a wizard wired into their legacy database. Five steps, straight into the company's SQL Server. It also carries a market-price column that scrapes the industry index, so a buyer sees where the price they just closed actually sits. - Stack: Next.js, SQL Server, SSH tunnel, Supabase - Status: replaced the spreadsheet ### DAOM Rugby One app for a rugby club: physical testing, wellness and the money side of the tour. Built module by module as the club asked for it: a live leaderboard for the testing days, a weekly wellness check the coaching staff reads, and the pot, the events and who owes what for the tour. - Stack: Next.js, Supabase, three modules - Status: 575 tests green - URL: ### Decision dashboard Listing and campaign data, reduced to the handful of numbers an agent decides with. Built for a RE/MAX agent who was reading three tools to answer one question. Two phases, both in production. - Stack: React, Meta Graph API, Recharts - Status: two phases shipped ### Legal automation A robot that pulls a law firm's case files from the insurance portal every day. It logs in, downloads the files and the paperwork, and hands the office a dashboard instead of a manual check every morning. Runs unattended on a VPS. - Stack: Python, Playwright, Hetzner VPS - Status: runs daily, unattended ## Experience — two things I built from nothing ### Co-founder, Prophet — December 2022 to present Prophet agentizes companies: their tasks become tools any AI agent can use. I run scoping, pricing, design and delivery, and I talk to the client myself. Everything above is mine end to end: the conversation, the interface, the database and the deploy. I productised one service, Demand-Driven Listings: instead of selling ad management to real estate agencies, the campaigns are read as market research, so the agency learns which properties, prices and neighbourhoods the demand reacts to. ### Founder, Odin Hood — August 2020 to 2026 My own clothing brand: product, costing, pricing, e-commerce and every campaign. Nothing was outsourced, so I learned what a decision costs before it reaches the customer. I ran Meta Ads, Google Ads, email and marketing automation myself — that is where the media buying came from. ## Toolkit - **Building:** Claude Code, Next.js, React, TypeScript, Python, Supabase and PostgreSQL, Vercel, n8n, Playwright - **Marketing:** Meta Ads, Google Ads, email and marketing automation, analytics, positioning - **Design:** Interface design and design systems, Figma, data visualisation, typeset PDF - **Education:** UADE — Bachelor's Degree in Marketing, 2024 - **Languages:** Spanish (native). English (advanced — Cambridge First, bilingual schooling) ## Contact - Email: - WhatsApp: [+54 9 11 6184 7541](https://wa.me/5491161847541) - LinkedIn: - Location: Buenos Aires, Argentina (UTC−3). Open to remote. ## For agents - [/llms.txt](https://bautistafilippini.com.ar/llms.txt) — index and when-to-use guidance - [/agents.md](https://bautistafilippini.com.ar/agents.md) — instruction file - [/sitemap.xml](https://bautistafilippini.com.ar/sitemap.xml) — every URL - MCP server: `https://bautistafilippini.com.ar/mcp` (Streamable HTTP, read-only, no auth) Every page is available as Markdown: add `.md` to the path or send `Accept: text/markdown`.