Coding agents and access

Claude Code integration

TesterArmy's hosted MCP server lets Claude Code create, run and read tests without leaving its own session.

START TESTING FOR FREE READ THE SETUP DOCS

[01] Session

From plain-language ask to verdict

Claude Code turns a plain-language request into a saved signup test. It reads the test-writing guide with get_guide, then calls create_test, run_test and get_run for the verdict.

Claude Code · testerarmy MCP

Create a test in the Web project that signs up a new user with a temporary email, verifies the welcome email arrives, and asserts the dashboard loads.

get_guide(how to write reliable test steps)guide read

create_test(saved test with plain-language steps)test saved in the Web project

run_test(remote run in TesterArmy cloud)run queued

get_run(status and step results)verdict passed

The test is saved in the Web project with plain-language steps for the signup, the welcome email and the dashboard. The run was queued in TesterArmy cloud and finished with a passing verdict.

Plugin install
The tester-army/cli repository is also a Claude Code plugin that installs the MCP server and the testerarmy-cli skill.
Short-lived token
The MCP server never sees the password, and Claude Code refreshes its short-lived access token automatically.
Credentials redacted
Stored login passwords never come back through a tool, and credential-bearing steps are redacted from the transcript.

[02] Setup

How it works

Setup is a single claude mcp add command, and a first reply of No authorization provided is expected until /mcp finishes the browser sign-in.

Add the testerarmy server

In a terminal run claude mcp add --transport http testerarmy https://tester.army/mcp, which registers the hosted server under the name testerarmy.

Sign in with /mcp

Run /mcp inside the Claude Code session to complete the sign-in, which uses OAuth 2.1 with dynamic client registration and finishes in the browser.

Ask Claude Code to test

Once signed in, describe the flow in plain language and Claude Code creates the saved test with create_test, queues it with run_test and polls get_run.

Read the setup docsThe docs cover the claude mcp add command, the /mcp sign-in, the agent's tools, security and troubleshooting.

See it on your own app

START TESTING FOR FREE

[03] Related

Works with

Codex and Cursor reach the same server through their own setup, the MCP server page covers every other client, and the GitHub App tests the pull requests Claude Code writes.

CodexCodex creates tests, runs saved groups and reads failures over the hosted MCP server.CursorCursor writes, runs and triages TesterArmy tests without leaving the editor.MCP serverClaude, Cursor, Codex and VS Code create tests, start runs and debug failures.GitHubThe GitHub App posts checks and comments on every pull request preview.

[04] FAQ

Questions

Once the hosted MCP server is added, Claude Code creates saved tests with plain-language steps, queues runs in TesterArmy cloud, polls the result with get_run and reads the transcript and telemetry of a failure.

Running /mcp inside a Claude Code session completes the TesterArmy sign-in over OAuth 2.1, so there is no key to create, paste or rotate. Key-based automation uses the ta CLI or the public API.

In Claude Code, /plugin marketplace add tester-army/cli followed by /plugin install testerarmy@testerarmy-agent-skills installs the MCP server and the testerarmy-cli skill together from the tester-army/cli repository.

The exploration agent reads the whole pull request description before planning. Paste the docs' excerpt into AGENTS.md or CLAUDE.md once, and every pull request Claude Code writes ends with instructions the planner can follow.

[05] Start in Claude Code

Connect Claude Code to TesterArmy

Add the server with one claude mcp add command, run /mcp, and sign in when the browser opens.