testerarmy (alias ta) is an agent-first control plane for the TesterArmy dashboard. Use it to manage projects, environments, credentials, and saved tests, and to queue remote runs that execute in TesterArmy cloud. It works standalone for interactive use or as a skill for coding agents like Claude Code, Codex, and OpenCode.
Your coding agent orchestrates dashboard-managed QA to validate changes - keeping the feedback loop tight without polluting main agent context.
Install
testerarmy and ta map to the same CLI.
Agent workflow
Agents can discover the full flow fromta --help. The usual path is:
ta tests run queues a run in TesterArmy cloud. For development targets, save a reachable preview or tunnel URL as a project environment and select it with --env.
If something in TesterArmy itself breaks, misleads, or is missing, the agent can tell us directly. Reports go to the TesterArmy team, not to your project:
--type is bug, docs, feature, or other (default). Run ta feedback --help for every field.
Agent skill (recommended)
If you use a coding agent, install the official skill for tighter integration:Authenticate
Runta auth to sign in through your browser:
--no-browser to skip it anywhere else.
To pick the workspace up front, or to switch later without signing in again:
--team takes the workspace’s slug, ID, or name.
Coding agents and other tools that run the CLI without an interactive terminal use --device. It prints the link and code as plain lines, then waits for someone to approve. Combine it with --team when the account has several workspaces:
ta signout ends the session and removes it from ~/.config/testerarmy/config.json.
Use an API key instead
For CI, or anywhere nobody can approve a browser sign-in, use an API key. Create one in the TesterArmy dashboard under Settings → API Keys; a key belongs to one workspace. Pass it directly, or runta auth --api-key without a value to paste it into a hidden prompt:
Verify
--json for machine-readable output:
Sign out
Connect your editor
Working in Cursor, VS Code, Windsurf, Claude Code, Codex, or another agent add-mcp supports? One command writes the hosted MCP server into every coding agent it finds, so your agent can create tests, start runs, and read failures without leaving the editor:Next steps
Local Development
Test dev changes through a preview or tunnel URL.
Agentic Usage
Give your coding agent the CLI and let it create projects, memories, tests, and runs.
Environments
Create stable staging or QA targets for remote runs.
Command Reference
Every command and its key flags.