Skip to content

Take the last manual step off your team

OpenAdapt enters approved information into the systems you already use, such as EMRs, payer portals, and desktop and Citrix apps, and checks that it saved. When something doesn't match, it stops and asks a person instead of guessing.

  • Evaluate it for one workflow (no install)

    See what changes in your team's day, what you control, where your data stays, and how a first project runs, in plain language.

  • Install and run it yourself

    The engine is free and open source (MIT). Install the Desktop app, or run pip install openadapt with Python 3.10 to 3.12, and try the tutorial on your own computer.

Watch a recorded run on OpenEMR with synthetic data

What happens in one run

  1. Someone on your team does the task once while OpenAdapt records it.
  2. OpenAdapt turns the recording into a fixed set of steps. Your team tests it on your own cases before it handles real work.
  3. For each item, it opens the record, checks that it's the right patient or account, and enters the approved information.
  4. It reads the saved record back through a second path, such as a report, an API, or a read-only login, and compares it with what it meant to enter.
  5. If everything matches, the item is done and checked, and you get a receipt. If something doesn't match, it stops and asks a person.

A normal run makes no AI model calls. For work with patient data, your screens and records stay on a computer you control. Read Security and where data stays for the details, or the glossary for the terms the technical pages use.

For developers, IT, and AI agent teams

The rest of this page uses the technical terms. OpenAdapt compiles a demonstration into a program that a person or an agent can invoke. Healthy runs make no model calls. Uncertainty escalates to a person.

  • An inspectable program

    The compiler turns the demonstrated actions into explicit steps, parameters, guards, and expected state changes. You can review the program before it runs.

    The demonstration compiler

  • Evidence for each consequential step

    A step can carry structural selectors, accessibility data, visual anchors, identity requirements, and a declared result check. The available evidence depends on the application surface.

    The capability ladder

  • A typed outcome and report

    VERIFIED means the declared result check passed. A run that cannot meet its identity, result, policy, or state requirements stops and records the reason.

    Run outcomes

The healthy path follows the approved program with no generative-model API calls. When the interface changes, OpenAdapt first uses the evidence retained from the demonstration. A configured model can propose a repair when policy permits. The repaired version passes the same review and qualification gates before promotion.

OpenAdapt separates action delivery from result verification. A save message on the acting screen does not prove that the intended record changed. A workflow can verify the result through an API, database, document store, read-only session, or persisted-state reacquisition.

Read about effect verification Read about identity checks

Where it runs

Surface Execution options Evidence available
Browser Local, customer-controlled, or managed for approved workflows DOM, accessibility, visual, geometry, and interaction evidence
Windows, macOS, Linux Local or customer-controlled Native accessibility data plus retained visual evidence
RDP, Citrix, VDI Local or customer-controlled Window-scoped pixels, keyboard, mouse, OCR, and visual anchors

Each workflow is qualified against its exact application, environment, identity rules, result verifier, and deployment boundary. The qualification evidence records the supported scope and its limits.

Sensitive runtime observations stay inside the declared execution boundary. Only an approved, sanitized derivative can cross that boundary.

Compare deployment options Review security and data handling

Inspect the evidence

The public references include the source demonstration, compiled replay, declared verifier, run counts, failure categories, and limits for each claim. The visual demo helps you understand a run. The result verifier and retained report determine its outcome.

Watch the real-application demo Review qualification evidence Read the benchmark methods

Check the current Production status

Connect an AI agent

OpenAdapt is the governed last mile an agent calls when the next write has no API. A person or a local agent they watch demonstrates. A named human admits the program. The calling agent operates it. The human returns for identity, effect, and judgment halts, then samples seals.

OpenAdapt supports automation teams, BPOs, service providers, and software companies that operate browser, native desktop, RDP, Citrix, or other virtual desktop applications.

What the calling agent may do / must not do

May: bind declared parameters, invoke a compiled program, read VERIFIED / HALTED / RECONCILIATION_REQUIRED, supply a missing declared parameter, retry a retryable transport failure, escalate.

Must not: summarize halt as success, resolve an identity or effect contradiction, teach by emitting guessed clicks, or be the sole source of a production demonstration.

Machine contract: agents.txt. Outcome vocab: Run outcomes.

Choose your next step

  • Evaluate one workflow

    What changes for your team, four worked examples, and how a first project runs. No install needed.

  • Install OpenAdapt

    Desktop or pip on this computer. Tutorial is optional.

  • Author a workflow

    A person or a local agent they watch records one browser workflow. A named human admits the compiled program and inspects the report.

  • Prepare a production workflow

    Define the identities, results, failure cases, policy, and deployment boundary for one exact workflow.