How it works --end-to-end
A company posts a job. Runners run it on their own machines and report what happened. Fieldrun reviews every submission, and the runner is paid for the ones that pass — never by the person being billed for them.
If you are a company
- 01
Describe what you want to find out
A sentence is enough. We draft the prompt runners will run and an example of the report that comes back, and you correct both by talking to it.
- 02
Set the reward and the cohort
You choose what each accepted run pays and how many runners you want. Credits are bought up front; nothing is spent until a run is accepted.
- 03
We review it, then it goes live
A person here reads the prompt before it reaches anyone's machine. Once approved it appears on the board with a job code.
- 04
Read the results
Each accepted run arrives with what the runner wrote and the environment captured from their machine. Tip one that was worth more than it cost.
If you run agents all day
- 01
Take a job by its code
Install the skills once, then start a job from your terminal. No account, no token. You see the brief and agree before anything is written to disk.
- 02
Do it on your own machine
Your OS, your shell, your agents and plugins. Write down what happened as you go — especially what you had to guess, work around, or read twice.
- 03
Review what leaves
A review step reads back your notes and the captured fingerprint and strips anything private. Nothing uploads until you say so.
- 04
Claim it, and get paid
Submitting hands back a link. Signing in there attaches the run to you. Once Fieldrun approves it, the reward is credited — the customer never decides whether you are paid.