Skip to content

CLI

This page ports upstream’s docs/reference/cli.mdx (tester-army/e2e, 17a2e46, #1711), for the commands the .NET port has. Only e2e explore is listed flag by flag; run e2e <command> --help for the others. The port has no e2e run, e2e list, or e2e upgrade: the tests run under dotnet test (see Testing frameworks), so the pages of upstream’s reference for those commands have no counterpart here.

Terminal window
e2e <command> [options]
Command What it does
e2e explore [goal] Runs the agent against the app toward a goal, without a test file. See e2e explore.
e2e mcp Serves a coding agent over MCP (stdio). See Your coding agent.
e2e doctor Checks that e2e explore and e2e mcp can start here. See e2e doctor.
e2e cache <ls|clear|stats> Reads the replay cache the runs write. See Caching.
e2e init [directory] Scaffolds a test project, installs the agent skill, and registers the MCP server.
e2e install [browser ...] Downloads browsers for the Playwright version E2E runs. Installs chromium when none is named.
e2e login, e2e logout, e2e models Sign in to a subscription provider, sign out, and list a provider’s models.
e2e guide [topic] Prints a guide of the installed skill.

e2e <command> --help lists the options of one command.

Terminal window
e2e explore
e2e explore "Explore checkout like a first-time buyer and report anything off"
e2e explore --target web --max-steps 4 --headed
e2e explore --changed-since origin/main
e2e explore "Hunt for broken forms and dead links" --video

Runs Explore.ExploreAsync from the project directory. The goal is Explore the app and find bugs when you give none.

Flag Meaning
--config <path> The config file. Default: the nearest e2e.config.json.
--target <id> The target to explore. Default: the config’s target.
--agent <name> The configured agent to explore as. Default: agents.default. Its model, system, tools, and context are reused. A custom Executor is replaced by the built-in agent, with a notice. An entry with no model is MODEL_UNAVAILABLE.
--changed-since <ref> Explores the change since the merge base of the ref and HEAD. The planner reads its commit messages and committed diff. With no goal, the goal is Explore what this change affects and find bugs in it, and the planner can skip a change no user can see; the run ends skipped and passes. An empty ref, one starting with -, or one git can’t measure from is INVALID_CHANGED_SINCE, exit 2. The report records the change as run.explore.change. See Explore a pull request’s change, and Security for what the model reads.
--max-steps <n> Exploration steps at most, an integer 1 through 100. Default: 8.
--timeout <ms> The run’s wall clock, an integer 180000 through 7200000. Default: 600000. The last minute is kept for the closing assessment.
--headed Shows the UI while the agent explores.
--output <dir> The output directory. Default: output in the config, else .e2e.
--debug Prints the phase timings and the agent step table to stderr (also E2E_DEBUG=1).
--ai-trace Records every model call to ai-trace-<process id>.json (also E2E_AI_TRACE=1).
--trace [mode], --video [mode] Which attempts keep a trace page or a video. A bare flag is on. The modes are off, on, retain-on-failure, on-first-retry, and on-all-retries; E2E_TRACE and E2E_VIDEO set them too. Both take the next word as the mode, so write --video=on "goal".

A flag out of range exits 2 and names the range. --session and --reporter exit 2: the port has no setup test to run for a session, and no reporters.

<output>/report.json gains run.explore: the goal, the budgets, why the run ended (finished, skipped, step-limit, time, stuck, or aborted), the assessment, the change with --changed-since (base and files), the steps, and the findings.

Result Status Exit code
Steps ran, no issue finding, and not every step was blocked passed 0
At least one issue finding failed 1
No step ran and nothing was found, unless the planner skipped the change blocked 1
With --changed-since, the planner skipped a change no user can see passed 0
Every step was blocked, with no issue finding blocked 1, 2, or 3: the class of the first blocker’s error code (see Errors)
Config, engine, or model provider error error 2 or 3
Ctrl-C 130