# RPA, AI agents, scripts or an API: which fits your data entry?

> Use an approved API whenever one does the job. Use RPA if you already run a platform with people to maintain it. Use an AI agent alone for varied, one-off tasks. For entries that repeat, give the agent OpenAdapt, which compiles the agent's run into a program that reads each save back and stops on a mismatch.

- Canonical: https://openadapt.ai/compare
- Updated: 2026-11-19

## At a glance

|  | An approved API | RPA platform | An AI agent alone | Your own script | OpenAdapt, alone or under your agent |
| --- | --- | --- | --- | --- | --- |
| Best for | Any task the API covers | Many automations on one platform | Varied or one-off tasks | A developer-owned task | Repeated entries with no usable API |
| Who builds it | Developers, with the vendor | Your RPA team | Anyone who can describe the task | A developer | We build it with you |
| AI model calls per run | None | Usually none | Every step | None unless added | 0 in our recorded runs |
| Proof the record saved | The API's reply | What the builder adds | Usually the model reading the screen | What the developer adds | A second read of the record, where available |
| When it isn't sure | Returns an error | Error handling the builder sets | The model decides | What the code says | Stops and asks a person |

OpenAdapt reads the saved record a second time because a screen can look right when the record is wrong. In our fault test, a check that looked only at the screen passed 54 of the 72 wrong records, and a full read of the database caught them all. [How the test worked](https://openadapt.ai/guides/know-rpa-bot-saved-the-record#numbers).

## Your agent alone and with OpenAdapt, timed

We timed OpenAdapt and Claude Sonnet 5.5 working alone as they entered the same new-patient referrals on an OpenEMR test copy, and they saved the same number correctly. OpenAdapt's program finished the median entry in 29.7 s, sooner than the agent, and its runs had no model fees. [The full comparison](https://openadapt.ai/compare/computer-use-agents#method) has the dates, the method and the counts.

## What OpenAdapt adds to your agent

Your agent can do a new task the first time, and OpenAdapt compiles the successful run into a program so that later runs make no model calls. Your 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.

### Technical details

The agent connection is the OpenAdapt MCP server, started with `openadapt agent serve` (launcher 1.17.0, openadapt-agent 2.2.1). Compiling an agent's run into a program is the JIT path in openadapt-flow 2.2.0 (openadapt_flow.jit), tested on a recorded Claude run against a synthetic app.

[The full table: your agent alone, or with OpenAdapt](https://openadapt.ai/compare/computer-use-agents#glance).

## Comparisons

- **[Your agent alone, or with OpenAdapt](https://openadapt.ai/compare/computer-use-agents)**: See what OpenAdapt adds to a computer-use agent in our measurements, including the tests where the agent alone did as well or better.
- **[OpenAdapt vs UiPath](https://openadapt.ai/compare/uipath)**: UiPath is a mature RPA platform, and this comparison explains when a narrower tool fits better.
- **[OpenAdapt vs Power Automate](https://openadapt.ai/compare/power-automate)**: Power Automate is strong inside Microsoft 365, and OpenAdapt covers a last step that is a screen with no connector.
- **[API or screen automation?](https://openadapt.ai/compare/api)**: Use the API whenever it covers the task.
- **[OpenAdapt vs browser agents](https://openadapt.ai/compare/browser-agents)**: Browser agents automate web tasks quickly with a model in the loop.
- **[OpenAdapt vs record-and-replay tools](https://openadapt.ai/compare/record-and-replay)**: Many tools can record a task, and OpenAdapt adds checks after the recording.
- **[OpenAdapt vs hand-written scripts](https://openadapt.ai/compare/hand-rolled-scripts)**: Scripts in Playwright, Selenium and AutoHotkey are free and precise, and all three are compared together.
- **[OpenAdapt vs Playwright](https://openadapt.ai/compare/playwright)**: Playwright is for developer-owned browser automation.
- **[OpenAdapt vs Selenium](https://openadapt.ai/compare/selenium)**: Selenium is the standards-based browser driver.
- **[OpenAdapt vs AutoHotkey](https://openadapt.ai/compare/autohotkey)**: AutoHotkey runs fast Windows macros on one machine.

## Questions

### Is OpenAdapt RPA?

It's a kind of screen automation, like RPA, built for a narrower job: a repeated entry where the saved record is read back and a mismatch stops. It doesn't offer a platform for hundreds of automations.

### Does OpenAdapt use AI?

A model can help build and repair a program. A normal run of the program doesn't call one, and none of OpenAdapt's recorded runs did.

### Can we use OpenAdapt with our RPA platform?

Yes. You can keep your platform for most work and use OpenAdapt for the few entries that need a read-back.

## Sources

1. 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.
2. Referral timing run, local OpenEMR 8.0.0.3 test copy, 14 to 15 Oct 2026 ([run data](https://github.com/OpenAdaptAI/OpenAdapt/tree/main/packages/flow/benchmark/openemr_referral_timing)): 8 synthetic referrals, OpenAdapt (a program compiled to find each field by the page's structure, 0 model calls) against Claude Sonnet 5.5 (claude-sonnet-5-5) as a computer-use agent in a capped harness, warm starts, interleaved blocks, each chart read back from the database. Blocks 2 to 10, normal referrals saved correctly: 62 runs each; median 29.7 s (OpenAdapt) and 36.5 s (agent); agent median $0.083 a run at list price. Both 62 of 63 normal referrals correct; on the possible duplicate, no new chart in 9 of 9 for both. Block 1 left out (CPU shared with another process); with it (blocks 1 to 10), 69 of 70 correct for both, 29.7 s and 37.6 s (n=69 each). The same task from the 10 Oct recording without the page's structure: 62 of 63, 38.6 s median, slower than the agent.
