AI Used-Car Auction Assistant
Budget: $250 – $750 USD
I’m building a personal AI agent that will take over the dull parts of used-car auction research and let me concentrate on the buys that really matter. Here is the experience I’m after:
• The agent logs into a couple websites and pulls the full market snapshot for every vehicle on my watch list.
• It then runs pricing and forecast analysis so I can instantly see potential margin, how the ask compares with current market averages, and short-term price movement expectations.
• Runs need to be both scheduled (cron-style) and on-demand, with all results stored so the agent can reference prior pulls and “remember” new tasks or tweaks I teach it. I’ll handle the task-teaching from the Google Chrome Claude extension.
• Voice access from my phone is essential. Think of a quick voice prompt (“check saved trucks”) and a spoken summary back to me, along with a link to the full report.
• The back end should run on my existing stack—either Vercel with GitHub or a DigitalOcean droplet—using Claude for the reasoning layer. All code and prompts must be easy for me to extend.
Deliverables
1. Authenticated web-scraping or API modules for each site I provide.
2. Data store and schema that keep historical pulls available to Claude memory.
3. Analysis service producing profitability and trend forecasts, returned in a clean JSON object.
4. Cron and manual trigger endpoints.
5. Voice interface (browser-based or mobile-friendly) wired to the agent.
6. Setup docs and a quick Loom or similar walkthrough so I can maintain and expand the system.
I’ll hand over logins, site lists, and any sample vehicles once you’re ready to dive in.
• The agent logs into a couple websites and pulls the full market snapshot for every vehicle on my watch list.
• It then runs pricing and forecast analysis so I can instantly see potential margin, how the ask compares with current market averages, and short-term price movement expectations.
• Runs need to be both scheduled (cron-style) and on-demand, with all results stored so the agent can reference prior pulls and “remember” new tasks or tweaks I teach it. I’ll handle the task-teaching from the Google Chrome Claude extension.
• Voice access from my phone is essential. Think of a quick voice prompt (“check saved trucks”) and a spoken summary back to me, along with a link to the full report.
• The back end should run on my existing stack—either Vercel with GitHub or a DigitalOcean droplet—using Claude for the reasoning layer. All code and prompts must be easy for me to extend.
Deliverables
1. Authenticated web-scraping or API modules for each site I provide.
2. Data store and schema that keep historical pulls available to Claude memory.
3. Analysis service producing profitability and trend forecasts, returned in a clean JSON object.
4. Cron and manual trigger endpoints.
5. Voice interface (browser-based or mobile-friendly) wired to the agent.
6. Setup docs and a quick Loom or similar walkthrough so I can maintain and expand the system.
I’ll hand over logins, site lists, and any sample vehicles once you’re ready to dive in.