Testing

Agency has a built-in test runner. See the testing guide for an overview of how Agency tests work; this page focuses on the CLI.

Running tests

agency test foo.test.json
agency test tests/

You can pass .test.json files or directories; directories are scanned recursively. With no arguments, the test runner uses the default test directory.

Options

• -p, --parallel <number> — number of test files to run in parallel.
• --coverage — collect step coverage and print a summary report after the run. See coverage.
• --accumulate — preserve any existing coverage data instead of cleaning it first (use with --coverage).

Generating fixtures

agency test fixtures foo.agency:nodeName

Run a node interactively, capture its behavior, and write a .test.json fixture next to the source file. The target is file.agency:nodeName. See the testing guide for the full fixture flow.

JS integration tests

agency test js tests/

Runs JavaScript integration tests (Agency code paired with .test.ts files that exercise it from TypeScript).

• -p, --parallel <number> — number of test directories to run in parallel.
• --coverage — collect step coverage and print a summary report after the run. See coverage.
• --accumulate — preserve any existing coverage data instead of cleaning it first (use with --coverage).

Evals

work in progress

agency test eval foo.agency:nodeName --args args.json

Runs an evaluation against a node. Arguments are read from a JSON file via --args. To resume an interrupted eval run, pass --results <path> pointing at the existing results file — the runner will pick up where it left off.