databrokerhub

Data that's cleared to sell, and ready to train on.

databrokerhub connects the companies and brokers who hold valuable data with the AI labs and model builders who need it. Every listing comes with a rights record, a scrubbed package and an audit trail, so a deal doesn't stall in diligence.

List a dataset Request data

Free to list. We take a success fee only when a sale closes. How fees and vetting work.

What a buyer receives 

      

      
Scrubbed in your browser by the same engine that builds every package. Edit the text to try it.

    

  


  

    

      
For sellers and brokers

      
    
        
  • List what you hold: the kind of data, volume, date range and how it was collected. Raw data stays with you until a license is signed.
    • 
        
  • Clear it once. We work through the clearance checklist with you and your counsel, and record the evidence for each answer.
    • 
        
  • See exactly what ships. Packages are scrubbed, sharded and hashed. You review the datasheet before any buyer does.
    • 
        
  • Set your terms: price, permitted uses, exclusivity windows, one-time or ongoing feed.
    • 
      

    

    

      
For buyers

      
    
        
  • Send a spec: domain, format, volume, freshness and budget. We match it against listings and sellers we're onboarding.
    • 
        
  • Read the datasheet first. Rights status, provenance, contents and known limitations, before you sign an NDA.
    • 
        
  • Verify what you receive. Every delivery has a manifest with a SHA-256 for each file and a hash-chained audit trail behind it.
    • 
        
  • Keep it fresh. Many sellers offer recurring feeds and exclusive windows.
    • 
      

    

  


  

    
What's in demand

    
Models can generate plenty of text. What they can't generate is a real record of real work. See current listings.

    

      
Work archives from wound-down companies
Years of chat, tickets, design docs and code review, packaged as threads and task trajectories.

      
Agent traces and failure logs
What agents tried inside real companies, where they went wrong and how a person fixed it.

      
Expert work products
Claims files, estimates, audits and filings, with the reasoning a professional used to get there.

      
Robot interventions
The moments a robot asked a human for help, with the sensor data around each one.

      
Evaluation sets
Hard questions with graded answers, written by people who do the work every day.

    

  


  

    
How a sale works

    
    
      
  1. List
    The seller describes the dataset. Buyers see a summary and a datasheet, never raw data.
    2. 
      
  2. Clear
    Every clearance item gets an answer and evidence, and counsel signs off.
    3. 
      
  3. Package
    Scrubbed, sharded and hashed into a package with a manifest and datasheet.
    4. 
      
  4. Match
    Qualified buyers review the datasheet and samples under NDA.
    5. 
      
  5. Close
    The license is signed and the package is delivered through databrokerhub. Our fee is due when the buyer pays. Details.
    6. 
    

  


  

    
The clearance standard

    
A package is marked cleared only when every item below has an answer and a pointer to evidence, and a named reviewer has signed off. Until then it's not for delivery.

    
    
This checklist organizes the review with counsel. It isn't legal advice and doesn't replace it.

  


  

    
Try it on a Slack export

    
Drop a Slack workspace export (the .zip from Settings, Import/Export Data). It's read, scrubbed and packaged inside this page. Nothing is uploaded.

    

      
      
      
      
or drop a .zip here