github.com/tester-army

Open source at TesterArmy

Everything we publish under the MIT license: tools for running QA with coding agents, testing APIs and reading agent traces, with more on the way. Built and maintained by the engineers behind TesterArmy, who also contribute to React Native core.

Explore projects Follow on GitHub

Claude Code v2.1.206

Welcome back Oskar!

Fable 5 with xhigh effort · Claude Max

TesterArmy

~/acme-app

Tips for getting started

Ask Claude to run the TesterArmy skill against a URL

What's new

Added directory path suggestions to /cd

Added a /doctor check that proposes trims

/release-notes for more

Test the checkout flow on staging before I merge #482

I'll run the checkout test on staging, sweep the API with Scout, then summarize the run's trace.

Run the checkout test on staging (completed)
Sweep the API with Scout (in progress)
Summarize the agent trace with unbox-ai (pending)

Bash(npx testerarmy tests run checkout --local)Result: ✓ PASS(ctrl+o to expand)

Result: ✓ PASS Target: https://staging.example.com All 6 steps passed.

Bash(npx @testerarmy/scout sweep --max-requests 25)Sweep complete: 25/25 probes run, 3 new findings (3 detected); stop=completed.(ctrl+o to expand)

[high] POST /orders — Accepts a negative quantity [medium] GET /orders/{id} — Returns another tenant's order [low] GET /products — No rate limit headers Next: `scout report` to compile, or `scout coverage` to see gaps.

Bash(npx unbox-ai summary trace.json)totals 21 generations in 8 segments, 310.8k in / 2.3k out, $0.07, 71.83s model time(ctrl+o to expand)

models google/gemini-3-flash, openai/gpt-5.6-luna caching 91% of input tokens were cached prefix (reported cache reads when available, else estimated from repeats)

·Thinking…(0s · ↑ 0 tokens · esc to interrupt)

xhigh · /effort

auto mode on (shift+tab to cycle) · ← for agents

OpenAI Codex(v0.132.0)

model:gpt-5.6-sol low/modelto changedirectory:~/acme-app

Test the checkout flow on staging before I merge #482

I'll run the checkout test on staging, sweep the API with Scout, then summarize the run's trace.

Ran npx testerarmy tests run checkout --local→ passed▸

Result: ✓ PASS
Target: https://staging.example.com
All 6 steps passed.

Ran npx @testerarmy/scout sweep --max-requests 25→ 3 findings▸

  [high] POST /orders — Accepts a negative quantity
  [medium] GET /orders/{id} — Returns another tenant's order
  [low] GET /products — No rate limit headers
Next: `scout report` to compile, or `scout coverage` to see gaps.

Ran npx unbox-ai summary trace.json→ $0.07▸

models    google/gemini-3-flash, openai/gpt-5.6-luna
caching   91% of input tokens were cached prefix (reported cache reads when available, else estimated from repeats)

Working(0s • esc to interrupt)

gpt-5.6-sol low · ~/acme-app

Oskar@testerarmy ~/acme-app

npx testerarmy tests run checkout --local

Result: ✓ PASS
Target: https://staging.example.com
All 6 steps passed.

npx @testerarmy/scout sweep --max-requests 25

  [high] POST /orders — Accepts a negative quantity
  [medium] GET /orders/{id} — Returns another tenant's order
  [low] GET /products — No rate limit headers
Next: `scout report` to compile, or `scout coverage` to see gaps.

npx unbox-ai summary trace.json

models    google/gemini-3-flash, openai/gpt-5.6-luna
caching   91% of input tokens were cached prefix (reported cache reads when available, else estimated from repeats)

npm downloads, all time: 37k
React Native core commits by our co-founder: 113
Merged pull requests by our co-founder: 1.5k+

Open source client

The same client our CI runs

The testerarmy package talks to the TesterArmy API from a terminal, a CI job or a coding agent. About ten thousand downloads a month.

Built for agents

Agent skills and safe defaults

One command installs the skill, and output is JSON on request. Scout blocks writes by default; unbox-ai never opens a server by accident.

Maintained by core contributors

Built by React Native contributors

Co-founder Oskar Kwasniewski has 113 commits in React Native core and maintains MiniSim, the simulator launcher for macOS.

[01] Projects

What TesterArmy publishes

Every package TesterArmy publishes on npm, with all-time downloads and GitHub stars read from the npm and GitHub APIs once a day.

TesterArmy's npm packages with all-time downloads and GitHub stars
Project	Downloads	Stars	Category
TesterArmy CLIAI QA agent that clicks through your website like a real user	31,254	47	Testing
ScoutAPI testing CLI for coding agents, driven by your OpenAPI spec	5,035	51	Agents
unbox-aiTurns an AI agent trace into a readable token and latency view	660	38	Agents

TesterArmy CLIAI QA agent that clicks through your website like a real user
Downloads31,254Stars47Testing
ScoutAPI testing CLI for coding agents, driven by your OpenAPI spec
Downloads5,035Stars51Agents
unbox-aiTurns an AI agent trace into a readable token and latency view
Downloads660Stars38Agents

[02] Get started

One command each

Every tool runs from npx and ships an agent skill. Scout and unbox-ai work without an account; the CLI needs a TesterArmy API key.

TesterArmy CLI

Runs your saved TesterArmy tests from a terminal, a CI job or a coding agent, remotely or in a local browser, and prints a pass or fail your agent can parse with --json.

npx testerarmy agent init

GitHub Docs

Scout

Parses your OpenAPI spec, runs instrumented requests inside a host allowlist, and records findings and coverage.

npx @testerarmy/scout agent init

GitHub npm

unbox-ai

Opens an agent trace as a token treemap, a latency waterfall and a deduplicated conversation view.

npx unbox-ai trace.json

[03] Contribute

Build with us in the open

Issues, pull requests and feature requests are welcome on every project. If you are wiring one of these tools into an agent workflow and get stuck, the maintainers are on Discord.

Join Discord Follow on GitHub