ecu.test agent

2026.1 FS2

Welcome

Time-consuming test preparation? We can cut it down drastically. With ecu.test agent, test case development is now faster than ever before, because our assistant generates directly executable test scripts in no time at all.

Features

Compatibility

The ecu.test agent is/will be compatible with multiple versions of ecu.test, including all feature sets of the latest version.

Compatibility Notice:
Please be aware, that the latest functionality documented in each release is only available when using that specific version of ecu.test. Please refer to the ecu.test documentation for detailed capabilities. Forwards compatibility is also not guaranteed, as new features may require changes in the agent.

Example

The agent is updated to support version 2025.4 FS2 (Feature Set 2), but you are using ecu.test 2025.4 FS1. The agent will still work, but you will not be able to use the new features introduced in FS1. The final version of the prior release, 2025.3, will of course also still work, but also not be able to use the new features of 2025.4 FS1.

Supported versions

Changelog

The agent version mentioned below corresponds directly to the latest supported ecu.test version of that release.

2026.1 FS2

  • Polling cycles can be defined for time options
  • Exit test steps can be generated
  • Comment test steps can be generated
  • Not yet supported test steps from reference implementations are represented as Todo steps

2025.4 Release

  • Units of mappings are now being used by the agent
  • User bug reports can now be generated with the ecu.test native logging tool

2025.4 RC

  • Support for ecu.test loop test steps
  • AI generated blocks are now retrievable via Object-API
  • Fix inconsistencies in "generated by agent" overlay in ecu.test
  • Fix JobMappings in the global mapping with autogenerated name prevent agent usage
  • Fix agent cannot generate, if current package has unsupported test steps
  • Beta feature: Trace Analysis step generation is now available. If you want to try this out, please contact support@tracetronic.com.
  • Supported trace steps:
    • Plot
    • Calculation
    • Trigger block

2025.4 FS3

  • Support for OAuth2 authentication with OpenAI LLM connections
  • Increased stability when generating test steps
  • Improved handling of signal names that are literally mentioned in a blocks description
  • Enabled configuration of multiple fallback LLMs

2025.4 FS2

  • Support now multiple custom headers for OpenAI LLM connections
  • Add support for Tool and Port Jobs directly from the current TBC
  • Add new version endpoint and print version on webpage entrypoint
  • Mark blocks as generated by the agent in ecu.test

2025.3 Release

  • Hidden files in ecu.test are excluded from context
  • Option for automatic reference implementation generation
  • Webpage entrypoint to display changelog
  • agent now supports OpenAI GPT-5

2025.3 FS7

  • Remove block test steps from generation
  • Whole testcase structure is now known to the agent
  • Add support for Tool and Port Jobs from global mapping

2025.3 FS5

  • Support for ecu.test block test steps
  • Sequential generation of test steps instead of parallel execution
  • Descriptions of global constants are now considered by the agent
  • More detailed progress dialog while generating

2025.3 FS4

  • Added AI transparency disclaimer
  • Hide api token in the ecu.test settings dialog by default
  • agent now also handles special Package parameters of type Enum, TextTable, Numeric or Boolean correctly

2025.3 FS3

  • ecu.test feature flag is no longer needed
  • Filter limits can now be defined independently for signals, Package mappings, lib Packages, variables and global constants
  • Further refined how agent processes Package parameters
  • Support for proxy settings in ecu.test to connect to server behind a corporate proxy
  • Support Azure OpenAI as LLM Provider

2025.3 FS2

  • agent now considers parameter descriptions and parameter type for Package calls
  • Enable multiline strings in test step parametrization
  • Any python expression can now be used in the supported test steps parametrization

2025.3 FS1

  • More robustness when generating multiple test steps
  • LLM support for GoogleCloud and AWS Bedrock
  • Add OpenIdConnect support in combination with OpenAI like LLMs
  • Version compatibility check between ecu.test and agent
  • Propagate LLM errors to user for better traceability
  • Custom instructions in ecu.test for project specific customizations
  • Prioritize reference implementation over mappings
  • agent shouldn't create inline comments for test steps anymore

2025.2 Release

  • First non official release
  • Test steps: read values, write values, If-then-else, multicheck, calculate, wait, call packages
  • Reference implementations: already implemented test steps can be reused and reparametrized