Your software depends on Stripe, Twilio, Clerk, and dozens of other services. WonderTwin lets you run them all on your laptop — offline, unlimited, no sandbox restrictions.
wt install stripe@latest twilio@latest clerk@latest
wt up
# ✓ stripe running on :4111
# ✓ twilio running on :4112
# ✓ clerk running on :4113
Born from running fleets of hundreds of integrations in production.
Mocks look right. Sandboxes work sometimes. But you keep finding the real bugs in production.
They don't track state, fire webhooks, or break the way real APIs break. You're testing against a fantasy.
Rate-limited. Incomplete. Can't simulate failures. Can't run offline. Can't scale.
Stripe had a blip. Twilio was slow. Clerk rotated something. Your CI is red and you did nothing wrong.
What happens when Stripe returns 429s during peak checkout? When Twilio drops webhooks? You find out when your customers find out.
A WonderTwin Stripe twin doesn't just return valid JSON. It tracks balances, creates transfers, fires webhooks, and maintains state across requests — just like the real Stripe API.
Built to pass compatibility tests against official SDKs.
# wondertwin.yaml
twins:
stripe:
version: "0.3.2"
port: 4111
seed: ./fixtures/stripe.json
twilio:
version: "latest"
port: 4112
clerk:
version: "~0.2"
port: 4113
wt up
# ✓ stripe running on :4111
# ✓ twilio running on :4112
# ✓ clerk running on :4113
stripe.SetBackend(
stripe.APIBackend,
&stripe.BackendConfig{
URL: "http://localhost:4111",
},
)
// Everything else stays the same
transfer, err := transfer.New(
&stripe.TransferParams{
Amount: stripe.Int64(1000),
Currency: stripe.String("usd"),
},
)
Your existing tests. Your existing SDK calls. Zero changes beyond the base URL.
Run your full dependency stack on your laptop. No network, no API keys, no rate limits. Local development, fully offline.
Deterministic integration tests. Single binaries, fast startup, zero external dependencies. Your CI goes green because your code works.
Inject real failure modes. What happens when Stripe returns 429s during peak? When Twilio drops 10% of webhooks? Find out before your customers do.
WonderTwin includes an MCP server so AI coding agents can discover, install, start, seed, and inspect twins natively. Unlimited API access for scenario testing at scale.
Point an AI coding agent at an API's public docs. Get a production-ready behavioral twin.
Twin generator skill. Feed an API's public documentation to any AI coding agent with the WonderTwin generator skill. It scaffolds the twin, wires the admin API, writes handlers, and produces a working binary.
Internal API twins. Your private APIs deserve twins too. Point the generator at your OpenAPI spec or internal docs and get a behavioral twin for your own services.
MCP server included. AI agents can discover, install, start, seed, and inspect twins natively through the WonderTwin MCP server.
View the generator skill →
All twins expose a standard admin API alongside their service-specific endpoints.
# Reset state between tests
curl -X POST localhost:4111/admin/reset
# Load seed data
curl -X POST localhost:4111/admin/seed -d @fixtures/stripe.json
# Inspect internal state
curl localhost:4111/admin/inspect/transfers
# Health check
curl localhost:4111/admin/healthWorks with any test framework. Go, Python, Node, Rust, Java — if it speaks HTTP, it works with WonderTwin.
Every twin is stateful, SDK-compatible, and free. Request one and we'll build it.
Accounts, Balance, Transfers, Payouts, Webhooks
PaymentsUsers, Sessions, Organizations, JWT validation
Auth & IdentityMessages, Verify (OTP send/check)
CommunicationsEmail send, delivery status
CommunicationsEvent capture, batch ingestion
AnalyticsStore API, product catalog
E-CommerceStore API, orders
E-CommerceLoyalty programs, points, rewards
Loyalty & RewardsReviews, loyalty, referrals
Loyalty & RewardsLoyalty programs, campaigns
Loyalty & RewardsCRM data, contacts, records
CRM & DataStore data, lead generation
CRM & DataLogo image retrieval
Utilitiestwin-{name}/
├── cmd/twin-{name}/main.go
├── internal/
│ ├── api/
│ │ ├── router.go
│ │ └── handlers_*.go
│ └── store/
│ ├── memory.go
│ └── types.go
├── go.mod
└── go.sumEvery twin follows the same structure. Standard layout. Standard admin API. Standard build pipeline.
Most twins can be generated by an AI coding agent in 2–4 hours from public SDK documentation.
Read the docs. Feed the SDK to your coding agent. Submit a PR.
Contributing Guide →
| WonderTwin | Mocks | Sandboxes | LocalStack | |
|---|---|---|---|---|
| Behavioral fidelity | ✓ State, logic, webhooks | ✗ Schema only | ~ Partial | ✓ AWS only |
| Offline / local | ✓ | ✓ | ✗ | ✓ |
| No rate limits | ✓ | ✓ | ✗ | ✓ |
| Chaos testing | ✓ | ✗ | ✗ | ✓ |
| SaaS coverage | ✓ 14+ services, growing weekly | ~ Manual | One vendor | ✗ AWS only |
| SDK compatible | ✓ | ✗ | ✓ | ✓ |
| Single binary | ✓ | ✗ | N/A | ✗ Docker |
Subscribe to fidelity.
We track official SDK releases and test twins against every new version. When something drifts, we know before you do.
API changelogs, deprecation notices, and breaking changes — all monitored and mapped to twin behavior.
When a twin needs updating, you get a PR with the fix, the test results, and the changelog reference. Merge and move on.
wt CLIwondertwin.yaml manifest
We're looking for design partners to shape WonderTwin Pro.
We'll be in touch soon with next steps.
# Install the CLI
go install github.com/wondertwin-ai/wondertwin/cmd/wt@latest
# brew install wondertwin-ai/tap/wt (coming soon)
# Add twins to your project
wt install stripe@latest twilio@latest
# Start everything
wt up
# Run your tests as usual
go test ./...