01
⏺
Record or write your flow
Use `ghostrun learn <url>` to open Chrome and record interactions, or write a `.flow.json` for API-only checks. Browser and API use the same flow model.
Browser · API · Load Testing
GhostRun CLI
Live npm package · MIT open source
Test everything.Entirely local.
Record browser flows or write API tests. Replay headlessly, run load tests with p95 stats, export to k6, capture screenshots, and use AI failure analysis only when you want it. The product is live now as `ghostrun-cli`.
0
Cloud calls
API + UI
Single workflow
k6
Export ready
ghostrun — bash
$ ghostrun learn https://myapp.com
RECORDING human flow — browser is live
navigate → https://myapp.com
click "Login"
fill #email = "user@example.com"
fill #password = "••••••••"
click "Sign in"
assert text: Dashboard
──────────────────────────────────────────
SAVED 5 actions recorded — human flow
$ ghostrun run "Login Flow"
Running: Login Flow
1 PASS Navigate to myapp.com 68ms
2 PASS Click "Login" 45ms
3 PASS Fill email 12ms
4 PASS Fill password 11ms
5 PASS Click "Sign in" 890ms
Passed in 1.1s
See it in action
Flow Library
Run Output
AI Chat
CLI Help
How it works
One tool.
Browser + API + Load.
No YAML. No extra framework to learn. Browser automation, API testing, and load testing all run through the same flow format.
02
▶
Run headlessly or under load
Replay with `ghostrun run <flow>`. API-only flows skip the browser. Use `ghostrun perf:run <flow>` for VU-based load testing with p50, p95, and p99 stats.
03
✦
Get plain-English failure analysis
When something breaks, GhostRun can use local Ollama or Claude to explain the failure, summarize the run, and help turn issues into next actions.
⚡Works without AI
Real browser recordingNo AI needed
Playwright headless replayNo AI needed
API testing with assertionsNo AI needed
Load testing with VU controlsNo AI needed
Flow schedulingNo AI needed
Screenshot diffingNo AI needed
Test suitesNo AI needed
k6 exportNo AI needed
✦Enhanced by AI
Failure analysis in plain EnglishOptional AI
Run summariesOptional AI
Natural-language to flow creationOptional AI
App exploration and crawl supportOptional AI
Code scan to draft flowsOptional AI
MCP tools for agent workflowsOptional AI
AI can default to local Ollama for private, offline-first usage. If neither Ollama nor Anthropic is configured, GhostRun still works as a non-AI CLI.
👻GhostRun Chat
Ask the built-in bot about flows, prior runs, or failed steps. It understands your local GhostRun workspace and stays offline-first with Ollama.
Ollama · offline-first
📊Monitor and diff
Schedule extraction flows, compare outputs across runs, and detect what changed without writing custom comparison code.
ghostrun monitor <flow>
🔍--visible mode
Agents can run headless, but humans can switch any flow into a visible browser session when debugging a flaky interaction.
ghostrun run <flow> --visible
🌐API testing
HTTP flows support assertions, JSONPath extraction, bearer auth, variable chaining, curl import, and OpenAPI-based starting points.
flow:from-curl · flow:from-spec
⚡Load testing plus k6 export
Turn any API flow into a load test, compare two performance runs, and export to k6 when you need a wider performance workflow.
perf:run · perf:compare · perf:export
🗂️Environments
Switch between dev, staging, and prod profiles without editing the flow itself. Environment values are injected before every run.
ghostrun env:use staging
Product vision
One tool. Four use cases.
01
Browser QA and E2E
Record real browser flows, replay headlessly in CI, and let GhostRun explain failures in language your team can act on.
ghostrun run → CI/CD
02
API testing
Model auth flows, assertions, extraction, and environment switching without having to split tooling between browser and backend checks.
ghostrun run → no browser
03
Load and performance
Turn API flows into repeatable load scenarios with latency percentiles, request throughput, and k6 export when you need to scale out.
perf:run · perf:export
04
Agent workflows
Connect GhostRun through MCP so Claude, Cursor, or other agents can create, run, inspect, and reuse flows autonomously.
MCP · Claude · Cursor
Template Store
Install a flow in
one command.
Production-ready starter flows for login, auth, health checks, checkout, CRUD, and load baselines. Use them as direct templates, not just examples.
login-basic
Navigate to login, submit credentials, and assert dashboard access
authstarter
ghostrun store install login-basic
signup-verify
Register a new account and verify the confirmation state
authsignup
ghostrun store install signup-verify
nav-smoke
Open key public pages and confirm they load without issues
smokeci
ghostrun store install nav-smoke
contact-form
Fill and submit a contact form, then assert success output
formsmarketing
ghostrun store install contact-form
checkout-guest
Add to cart and walk through guest checkout assertions
ecommercecheckout
ghostrun store install checkout-guest
api-health
Hit health endpoints and assert healthy responses
apimonitoring
ghostrun store install api-health
search-results
Run a search flow and verify results render correctly
searchsmoke
ghostrun store install search-results
password-reset
Request password reset and verify the confirmation step
authemail
ghostrun store install password-reset
api-auth-flow
Authenticate, extract token, and call a protected endpoint
apiauth
ghostrun store install api-auth-flow
api-crud
Create, fetch, update, and delete a resource in one flow
apici
ghostrun store install api-crud
load-baseline
Establish a p95 and RPS baseline against an API endpoint
perfmonitoring
ghostrun store install load-baseline
For AI agents
Flows know
who created them.
GhostRun can track whether a flow was recorded by a human or created by an agent. That makes it easier to let agents generate workflows while still keeping a trust boundary around what gets promoted.
A human can record a login once and let agents reuse it with environment variables. An agent can draft a flow from natural language or code context, then a human can verify and keep it.
👤
Human-recorded flows
Captured in a real browser by a person. Strong for verified interaction paths and trusted baseline flows.
createdBy: human
🤖
Agent-generated flows
Generated from natural language, code scanning, or exploration. Useful for drafting flows fast, then reviewing or promoting them.
createdBy: agent
Claude Desktop — MCP integration
tools: list_flows, run_flow, get_run_result,
create_flow, get_status, delete_flow
create_flow, get_status, delete_flow
Get started
Everything local.
Nothing to trust.
Your flows, screenshots, and run history stay on your machine. The installable package is `ghostrun-cli`; the command you run is `ghostrun`.
Install
$ npm install -g ghostrun-cli
First commands
$ ghostrun init # setup wizard
$ ghostrun chat # ask the bot anything
npm install -g ghostrun-cli→ghostrun learn https://yourapp.com→ghostrun run <flow-name>