OpenAdaptLoan application entry, checked against the stored application
OpenAdapt types a loan application into the origination system as a program, then checks the stored application through the system's own data. On Frappe Lending, an open-source loan system, 6 of 6 synthetic applications were checked, 3 of them after a cosmetic screen change, and 5 of 5 injected faults stopped before saving.
Updated 2026-10-18
What's been tested
Frappe Lending 16.2.0, a local test copy, synthetic applicants, 16 Jul 2026. Each run had to create exactly one loan application. An API read, a database read and an audit of which tables changed all had to agree.
| Case | Result |
|---|---|
| Normal runs | 3 of 3 Done and checked |
| After a cosmetic screen change | 3 of 3 found the fields again and were checked |
| Injected faults | 5 of 5 stopped before saving |
| AI model calls | 0 |
In a separate test on 21 Jul 2026, a paid computer-use agent also made 6 of 6 correct entries on Frappe Lending; 5 of the 6 runs ended cleanly, and in the other the agent hit its spending cap after the entry was saved. The two weren't run on the same baseline.
In practice
Frappe Lending has a full API, so in practice you'd use the API there. The screen program is for systems without one.
Related: Lending · when the screen changes
Sources
- OpenAdapt on Frappe Lending 16.2.0 (open source), 16 Jul 2026, synthetic applicants: 6 of 6 loan applications checked through the API, a database read and a table-change audit (3 normal runs, 3 after a cosmetic screen change); 5 of 5 injected faults stopped before saving.
- OpenAdapt engine benchmark, agent arm, 21 Jul 2026 (README): a paid computer-use agent, Claude Sonnet 5 (claude-sonnet-5), on local test copies with synthetic data, each trial scored by a database read. openIMIS claim entry: 3 of 3 correct, $0.48 a run at list price, 69.8 s mean. Frappe Lending: 6 of 6 correct writes, 5 of 6 clean. Local OpenEMR: 0 of 6 (each trial used up its action budget before saving). Small samples; run on fresh baselines, not matched against OpenAdapt runs.
See it on your process
Bring one task your team types every day. We'll tell you whether it fits, and what a test on your own system would cover.