Ask in plain words
Describe the flow as you would to a colleague and the agent takes it from there.
[02] From prompt to passing run
From prompt
to passing run
Your agent connects over MCP, writes a test in plain language, runs it in the cloud and reads the report.
The agent runs the test
A Claude Code session calls create_test and run_test over MCP, then get_run_report returns the transcript with every step and its result.
create_test
The agent turns the request into a saved test with named steps on the platform you choose.
run_test
The test runs in the cloud against the environment the agent names, such as Staging.
get_run_report
The report comes back into the session with each step, its result and the run duration.
Connect your agent and run the first test
Add the MCP server, paste the prompt and read the first result inside your editor.
[04] Three agents, one endpoint +
Three agents, one endpoint
Claude Code, Codex and Cursor connect to the same MCP server with one command, and any other agent can drive the ta CLI from a shell.
Connect once
Add the TesterArmy MCP server with one command, then pick the agent you work in. Tests and results are shared across all three.
Claude Code
Add the server with one claude mcp add command and the session picks up the TesterArmy tools.
Codex
Codex reaches the same MCP server and calls the same tools from its own session.
Cursor
Cursor Agent connects over MCP as well, so the tests it writes land in the same project.
CLI for any agent
Any agent with a shell installs the ta CLI, authenticates with an API key and runs tests.
[05] Explore the platform +
Run at every step of shipping
The same tests run on every pull request, from your CI job, on a schedule against Production, or by hand. Every run lands in one Results table.
FAQ
Frequently asked questions
What is the TesterArmy MCP server?
It is a hosted Model Context Protocol server at https://tester.army/mcp. It exposes projects, environments, tests, groups and runs to your coding agent, so the agent can create tests, run them and inspect results from your editor.
How does the agent authenticate?
The hosted MCP server uses an OAuth sign-in with your TesterArmy account. The CLI supports authentication with a TesterArmy API key. Access stays limited to the teams, projects and resources available to that identity.
Which coding agents can use it?
Claude Code, Codex, Cursor and other clients that support remote HTTP MCP can connect. An agent with a shell can also use the TesterArmy CLI to work with the same saved projects and tests.
Does this replace Pull Request Testing?
They work together. An agent can test a change while you develop it. Pull Request Testing then runs on the preview deployment and reports on the pull request before merge. Both use the same projects, tests and Results table.
Should the agent run tests locally or in the cloud?
TesterArmy executes tests in the cloud. To test development changes, use a reachable preview deployment or a tunnel URL and select that saved environment. The cloud browser cannot open your machine's localhost address directly.
Do runs from the CLI or MCP cost extra?
Runs follow your team's existing TesterArmy plan limits and usage. The execution path is the same whether a run starts from the dashboard, CLI or MCP.
Can the agent run a whole group of tests?
Yes. The MCP server and CLI can queue a saved test group against a selected environment. The agent can inspect each run's results, messages and artifacts afterward.
Can non-technical teammates still use the dashboard?
Yes. Tests created by an agent are saved in the same project, and every run appears in Results. Teammates can read and edit the plain-language steps, run tests and inspect recordings from the dashboard.
