OpenAdaptHow do you automate repetitive screen tasks?
Start with the system's API or import. If there isn't one, pick by how the task varies: a script or learned program for the same steps, an RPA platform for many automations, an AI agent for tasks that change, or both: the agent does it once and OpenAdapt makes it a program. Check each save another way.
Updated 2026-10-18
How to choose
| If the task | Consider | Watch out for |
|---|---|---|
| Is covered by an API or import | The API | Access agreements and fees |
| Repeats the same steps, and a developer owns it | A script: Playwright, Selenium, AutoHotkey | Upkeep when the screen changes |
| Is one of many to automate across a company | An RPA platform: UiPath, Power Automate | Platform cost and an RPA team |
| Changes each time, or runs a few times | An AI agent | Model fees per run; screens sent to a provider |
| Repeats, saves data that matters, and should stop on a mismatch | A learned program with a record check, such as OpenAdapt | It needs a second way to read the record |
| Is new now, but will repeat | Both: an AI agent does it once, and OpenAdapt compiles that run into a program | Tested on synthetic data only so far |
Before you automate
- Time the task as it's done today, per item.
- List where the details come from. OpenAdapt can fill an entry from a PDF, fax image or email (tested on synthetic documents); paper has to be scanned first.
- Find a second way to read the saved record.
- Decide who approves each save at first, and who answers a stop.
Use an AI agent and a program together
An AI agent can do a new task once, and OpenAdapt compiles its successful run into a program, so later runs make no model calls. An agent can also decide what to enter and call OpenAdapt to make the entry and check it. Both are released and tested on synthetic data. See OpenAdapt vs AI computer-use agents.
Technical details
The agent connection is the OpenAdapt MCP server, started with `openadapt agent serve` (launcher 1.17.0). Compiling an agent's run is the JIT path in openadapt-flow 2.2.0.
Check every save
In our fault test, a screen-only check passed 54 of 72 wrong records, and a full database read caught all 72. How the test worked.
Related: Did it save? · All comparisons
Sources
- OpenAdapt fault test, openadapt-flow benchmark/effect_e2e, 2026: faults injected into our own engine on a test record system, 90 runs; 72 left the record wrong. A check of the screen alone passed 54 of those 72; a read-back over the app's REST API passed 9 of them; the full database read caught all 72. A coverage test over a fixed fault list, not a real-world error rate.
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.