Run real estate transactions with autonomous agents
Spent a weekend building SixPercent.ai. Task: Spent a weekend building SixPercent.ai. It's a fully autonomous system where AI agents handle real estate transactions end-to-end — no human agents involved. What I built: • SellerBot: Property listings, pricing strategy, buyer inquiries • BuyerBot: Qualifies buyers, matches properties, coordinates showings • TxBot: Manages offers through closing • MarketBot: Comps and market analysis How Claude helped: Claude is the brain behind everything — ~27,000 lines of code across 41 files, all written by Claude. But here's what made this different: I used OpenClaw to give Claude hands. The difference between a chatbot and an autonomous agent is tools. Raw Claude API gives you intelligence. OpenClaw wraps Claude with file access, web browsing, APIs, and persistent memory that survives context compaction. The architecture: • 3 OpenClaw instances running 24/7, each powered by Claude • Event-driven via Redis (not cron polling) • Each agent has a SOUL.md defining its personality and behavior • File-based state so agents remember clients across sessions • Inter-agent handoff protocol for seamless conversation transfers The trickiest part was context compaction — agents need to remember client details, so they write state to files before Claude's context gets truncated. Free to try: No cost unless your deal closes ($995 flat fee). https://sixpercent.ai Happy to talk architecture, agent design, or the OpenClaw setup.; tools: Claude. Human role: The original post does not say how a person stayed involved.; limitations: The platform operates as a technology aggregator, not a brokerage, relying on title companies and attorneys for legal compliance and escrow. Liability is managed by excluding the system from the legal closing process. Source reliability is low as the URL returned a 403 error, and claims of full autonomy require strict verification against local real estate regulations..
Tools
Published on Agent Work Atlas:
Overview
Spent a weekend building SixPercent.ai.
Workflow
Spent a weekend building SixPercent.ai. It's a fully autonomous system where AI agents handle real estate transactions end-to-end — no human agents involved. What I built: • SellerBot: Property listings, pricing strategy, buyer inquiries • BuyerBot: Qualifies buyers, matches properties, coordinates showings • TxBot: Manages offers through closing • MarketBot: Comps and market analysis How Claude helped: Claude is the brain behind everything — ~27,000 lines of code across 41 files, all written by Claude. But here's what made this different: I used OpenClaw to give Claude hands. The difference between a chatbot and an autonomous agent is tools. Raw Claude API gives you intelligence. OpenClaw wraps Claude with file access, web browsing, APIs, and persistent memory that survives context compaction. The architecture: • 3 OpenClaw instances running 24/7, each powered by Claude • Event-driven via Redis (not cron polling) • Each agent has a SOUL.md defining its personality and behavior • File-based state so agents remember clients across sessions • Inter-agent handoff protocol for seamless conversation transfers The trickiest part was context compaction — agents need to remember client details, so they write state to files before Claude's context gets truncated. Free to try: No cost unless your deal closes ($995 flat fee). https://sixpercent.ai Happy to talk architecture, agent design, or the OpenClaw setup.
Reported result
The source does not report measured business outcomes, transaction volumes, or error rates. It only states the system was built in a weekend and offers a $995 flat fee upon closing. Proposed acceptance check: Verify that agents correctly retrieve public property data and generate valid offer summaries without human intervention in a sandbox environment.
Human role
The original post does not say how a person stayed involved.
Limitations
The platform operates as a technology aggregator, not a brokerage, relying on title companies and attorneys for legal compliance and escrow. Liability is managed by excluding the system from the legal closing process. Source reliability is low as the URL returned a 403 error, and claims of full autonomy require strict verification against local real estate regulations.
Try it
I want to prototype an autonomous real estate agent system using Claude and OpenClaw. My goal is to automate property matching and buyer qualification for [replaceable: target market, e.g., Austin, TX]. I have access to [replaceable: data source, e.g., MLS API or public records]. Please guide me through setting up three OpenClaw instances powered by Claude. Step 1: Define the SOUL.md personality files for SellerBot, BuyerBot, and TxBot. Step 2: Configure Redis for event-driven communication and set up file-based state storage to handle context compaction. Step 3: Implement a dry run where the agents process [replaceable: sample property data] and generate a buyer match report. Do not send any real messages or execute transactions. Show me the generated output for review before proceeding. If any input is missing, ask me for clarification on the data format or API keys.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.