Use cases#

TestZeus plugs into your workflow four ways. They all talk to the same platform (https://prod.testzeus.app/api) — pick the one that fits what you’re doing, and mix them freely.

Which one should I use?#

CLI — pip install testzeus-cli

A direct, scriptable handle on TestZeus — nothing to write, and it exits with a status code CI can gate on.

Reach for it when:

  • Quick one-off checks from your terminal

  • Shell scripts, Makefiles & cron jobs

  • Gating a deploy from a CI pipeline

Example: run a smoke suite in a GitHub Action and fail the build the moment a test goes red.

testzeus test-run-group execute-and-monitor \
  --name "Smoke" --tags smoke
CLI reference
Python SDK — pip install testzeus-sdk

A typed async client for when testing has to live inside your own product or service logic.

Reach for it when:

  • Embedding TestZeus into your app or backend

  • Custom automation & internal tools

  • Reading results to act on programmatically

Example: when a PR merges, your backend triggers a run and writes the pass/fail back to your own dashboard.

async with TestZeusClient() as c:
    run = await c.test_run_groups \
        .create_and_execute(name="ci", test_ids=ids)
SDK reference
MCP server — native tools for AI clients

Exposes TestZeus as native tools so a chat client can run and inspect tests right in the conversation.

Reach for it when:

  • You live in Claude Desktop or Cursor

  • Ad-hoc, conversational testing

  • No terminal or code in the loop

Example: ask Claude Desktop “run the login test and show me the failures” — it calls run_test and reports back.

claude mcp add testzeus -- testzeus-mcp-server
MCP Tools
Skills — for coding agents

Plain-English playbooks a coding agent loads on demand to author and run whole suites hands-off.

Reach for it when:

  • Inside a coding agent (Claude Code, Codex)

  • Authoring tests from a description

  • End-to-end flows you don’t want to babysit

Example: “Create a checkout regression suite and run it in staging” — the agent writes the Gherkin and runs it.

# in your coding agent
"create a login test and run it in staging"
Skills

How they fit together#

The four surfaces are two layers over the same API:

  • Skills + MCP are the agent-facing layer — natural language in, tests run, results back.

  • CLI + SDK are the direct/programmatic layer — for a human in a terminal or code in an app.

Because they share one backend, you can combine them: author tests with skills, then gate deploys with the CLI in CI.

Common use cases#

I want to…

Use

Gate a deploy in CI

CLI in your pipeline

Author tests in plain English

Skills in your coding agent

Let Claude Desktop / Cursor drive testing

MCP server

Embed testing inside my own product

Python SDK