Buyer · AI data company (production traces)

SF Data: what they pay for company data

Short answer

SF Data, which also calls itself San Francisco Data, works with companies that run AI products to turn their production traces, the records of AI working on real tasks, into training data, and says the source company can share in the proceeds. It says it is starting with coding agents. It publishes no price, terms or program page, and the only contact on its site is a founder's email address.

Website
Payment
Not published
Speed
Not published
Minimum
Not published
Fees to you
Not published
Referral reward
Not published
Kind
Buyer · AI data company (production traces)

What does SF Data buy?

  • Production traces from AI products: the user's request, the context the model had, its tool calls and later feedback
  • Coding agent sessions, where it says it is starting

How it works

As SF Data describes it on its own pages.

  1. SF Data's site publishes no process. It lists one contact, a founder's email address.

What we know about SF Data

Each fact links to its source below.

  • SF Data defines a production trace as the chronology of AI working on a real task, including the user's request, the context available to the model, its tool calls and later feedback.1
  • It says companies that operate AI products can share in the proceeds when frontier labs use their traces, without building a frontier model themselves. It describes its goal as a "Glencore for production traces": gathering fragmented supply, processing it and selling to concentrated buyers.1
  • SF Data says it works closely with each partner rather than at arm's length, and that this sometimes means designing a new pricing model for a kind of data no one has sold before. It publishes no price or terms.1
  • It says production sessions often contain customer information and proprietary material, so permissions and filtering decide which training tasks it can build.1
  • SF Data says most economically valuable traces today come from coding agents, and that this is where it is beginning.1
  • Frontier List lists SF Data as a team of 1 to 10 people in San Francisco, founded by Brandon Guo and Faiz Siddiqi.3
  • In an October 2026 informal poll of 36 people in AI by investor Chris Barber, 5 named SF Data as a company to talk to about selling a private dataset, such as a shut-down startup's data. Voters could not pick a company they work at or founded. Barber discloses that he is a small investor in SF Data.23

Is it a good fit?

Good fit

  • Companies that run AI agent products, especially coding agents, and keep session logs
  • Product companies that want a recurring share of proceeds rather than a one-time sale

Worth checking

  • No price, payment terms or process is published. Ask how proceeds are shared and when you are paid.
  • Production traces come from your users' sessions. Check that your terms of service and customer contracts allow you to license them before any conversation about samples.
  • Ask what customer information is filtered out, who does it, and who receives the resulting training data.
  • Ask whether the license is exclusive and how long it runs.
Questions owners ask

SF Data FAQ

What does SF Data buy?

SF Data works with companies that run AI products to license their production traces: the request, the model's context, its tool calls and feedback. It says it is starting with coding agents (sfdata.com, checked October 2026).

How much does SF Data pay?

SF Data does not publish a price (October 2026). It says the source company can share in the proceeds and that it sometimes designs new pricing models for each partner.

Does SF Data buy records from companies that do not run AI products?

Its site describes only production traces from AI products, so a company without one is not its stated focus. Buyers of general company records are listed in our buyer index.

Alternatives to SF Data

Other companies that pay for, broker or resell the same kinds of records. Sorted by name.

BuyerAfterQueryAI data company that licenses records of how teams work, from email and chat to CRM and finance, to train and evaluate frontier AI.Price not publishedAbout AfterQueryBrokerThat's usData LicensorsGets owners of 20 to 500 person companies written offers from several AI buyers for a de-identified copy of their records. Free for owners.Price not publishedAbout Data LicensorsBuyermicro1AI data company that licenses operating records from established companies to train and evaluate AI models.$100K to $2M+About micro1BuyerOoak DataParis-based AI research lab that licenses company records, anonymizes them and turns them into training environments for AI agents.$20K to $1MAbout Ooak DataBuyerPatronus AIAI company that licenses private codebases and company records from their owners to build training environments for AI agents.Code: $100 to $200 per million tokens; records: not publishedAbout Patronus AI

Sources

Numbers in the text link here. See every player in the buyer index and reported deals in the deal ledger.

  1. Production traces and the future of dataSF Data · Publish date not shown · Accessed 29 Oct 2026
  2. Post by Chris Barber: who to talk to if you're selling a private datasetX (@chrisbarber) · Published 29 Oct 2026 · Accessed 29 Oct 2026
  3. Private dataset buyersFrontier List · Publish date not shown · Accessed 29 Oct 2026

Some companies listed here may pay us a referral fee. This page is written from public sources and was checked on 29 Oct 2026. Something wrong or out of date? Email hello@datalicensors.com and we will correct it.

See what your records could earn

Six questions, about a minute. No upload, no system access.

Get my offers Book a 20-minute call