Skip to content

The OpenAdapt platform

From a demonstrated task to a checked result.

Create repeatable GUI workflows, run them within an agreed policy, and inspect what happened. Use OpenAdapt yourself, or have our team implement the workflow with you.

How it fits together

One workflow. Clear responsibilities.

Flow supplies the compiler and runtime. Desktop and Cloud give your team the interfaces to create and operate the workflow. Agent connects local MCP clients and Agent Skills. Execute provides the service integration path for applications and agents.

  1. 01

    Demonstrate

    Show the task in the application where it happens.

    Desktop or Flow
  2. 02

    Compile and qualify

    Review the program and define the checks for the intended result.

    Flow
  3. 03

    Run within policy

    Use the approved workflow in its agreed execution environment.

    Desktop, Agent, Cloud, or Execute
  4. 04

    Inspect the outcome

    Review the evidence and resolve any run that needs attention.

    Run reports and Seals

Seals provide signed execution outcomes. Evals is a separate toolkit for measuring behavior in benchmark environments; you don't need it to run a workflow.

Explore the software

Choose the part you need.

Find the supporting tools.

Install the OpenAdapt CLI, inspect recordings with Viewer, redact artifacts with Privacy, or work with the shared schemas. The software directory also covers research packages and the tools our team uses.

Get started

Bring a workflow or start building.

Let us deliver the workflow.

We define the scope and acceptance criteria, connect the systems, and prepare the team to operate it. Maintenance has an agreed owner and scope.

Explore implementation services

Use the software directly.

Start with a local demonstration. Review the compiled program, run the tutorial, and choose the deployment boundary for your own application.

Open the getting started guide