Local-first API client + mock server
StubHouse is the desktop API client that takes mocking seriously. Native, offline, file-based — built for the developer who has to ship before the API exists.
Available for macOS, Linux, and Windows.
Response
// Open workspace
One workspace for real requests, local mock rules, scenarios, and recorded fixtures.
Two tools. One binary.
Other API clients bolt on mocks as a paid afterthought or a separate process. StubHouse was built around the assumption that you spend half your day calling APIs and the other half pretending an API exists. Both deserve a real tool.
200 OK · 132 ms
{
"items": [{ "id": "u1", "name": "Ada" }]
}Scenario dropdowns live next to the routes they own.
Source of truth
No proprietary database. No cloud account required. A workspace is a `.stubhouse/` directory of YAML files: requests, mock rules, environments, scripts. Commit it to git. Diff it in PRs. Move it between machines with `cp -r`. The tool builds on top of files; the files outlive the tool.
id: req_get_user
name: "Get User by ID"
description: |
Retrieves a single user by their unique ID.
method: GET
url: "{{base_url}}/users/{{user_id}}"
headers:
- key: Authorization
value: "Bearer {{auth_token}}"
enabled: true
auth:
type: bearer
token: "{{auth_token}}"
tags:
- users
- readThe mock server
Switch your mock from `success` to `not_found` to `server_error` without restarting anything. From the UI, from the CLI, or from your test runner via the control API.
{
"id": "{{params.id}}",
"name": "Alice Nguyen",
"email": "alice@example.com"
}A mock that behaves like an API. POST creates. PUT updates. DELETE removes. GET returns what you wrote. All in memory, all reset on demand.
mock_resources:
- path: /users
id_field: id
seed_file: ./fixtures/users.yaml
auto_crud: truecurl -X POST http://127.0.0.1:4000/users -d '{"name":"Ada"}'
curl http://127.0.0.1:4000/users/usr_1
curl -X DELETE http://127.0.0.1:4000/users/usr_1Point StubHouse at a real API. Make some calls. Save them. Replay them offline. Forever.
Traffic passes through to the real service while rules are captured.
Chaos, on demand
Connection resets. Timeouts. Slow responses. Partial bodies. Random 5xx at a configurable rate. Every fault is a checkbox. Your retry logic finally has something to retry against.
connection_reset
TCP reset mid-response — exercise reconnect paths.
fault: connection_reset
timeout
No reply until the client gives up.
fault: timeout
slow_response
Fixed or random tail latency.
delay_ms: type: random min: 800 max: 3000
partial_body
Truncated payloads for parser hardening.
fault: partial_body
random_5xx
Rare server errors on demand.
fault: type: random_5xx probability: 0.12
passthrough
Forward to upstream when you need a real edge.
passthrough: true
Automation, sandboxed
StubHouse uses Rhai — a sandboxed scripting language designed for embedding in Rust. Pre-request scripts. Post-response assertions. Mock rule conditions. All run in a sandbox with no filesystem, no network, no escape. No 50MB V8 binary in your dock.
test("Status is 200") { response.status == 200 }
test("Body has user field") { response.json()["user"] != null }
test("Response under 500ms") { response.time_ms < 500 }Status is 200 · pass · 4 ms
Body has user field · pass · 1 ms
Response under 500ms · pass
It's a binary
stubhouse serve and ship.The same engine that runs in the desktop app runs from the command line. Spin up your mocks in CI. Run your assertion suite against them. Get JUnit XML out the other side. Ten kilobytes of YAML replaces a thousand lines of `nock`.
# CI: serve mocks, run tests, switch scenario
stubhouse serve . --port 4000 --env test &
MOCK_PID=$!
npm test
stubhouse scenario set mock_payment checkout_failure
npm test -- --grep "error handling"
kill $MOCK_PIDHonestly
A direct comparison with the tools you already know. We ship the embedded mock as part of the product, not as an upsell.
| StubHouse | Postman | Insomnia | Bruno | Yaak | |
|---|---|---|---|---|---|
| Local-first (no account) | ✓ | ✕ | ✓ | ✓ | ✓ |
| Files as source of truth | ✓ | ✕ | ✕ | ✓ | ✓ |
| Embedded mock server | ✓ | ◐ | ✕ | ✕ | ✕ |
| Stateful mocks | ✓ | ✕ | ✕ | ✕ | ✕ |
| Scenario switching | ✓ | ✕ | ✕ | ✕ | ✕ |
| Recording mode | ✓ | ◐ | ✕ | ✕ | ✕ |
| Fault injection | ✓ | ✕ | ✕ | ✕ | ✕ |
| Mock control API | ✓ | ✕ | ✕ | ✕ | ✕ |
| CI / headless mode | ✓ | ✓ | ✕ | ✓ | ✕ |
| OpenAPI schema validation | ✓ | ✓ | ✓ | ✕ | ✓ |
| Native binary (no Electron) | ✓ | ✕ | ✕ | ✓ | ✓ |
| Scripting language | Rhai | JS (V8) | JS (V8) | JS (V8) | — |
| Free and open source | ✓ | ✕ | ✓ | ✓ | ✓ |
Sourced from public information as of 2026-05-08. Updated when competitors ship. Corrections welcome — open a PR.
Under the hood
Tauri 2 shell. Rust core. `hyper` HTTP runtime. `rustls` for TLS. Embedded mock server runs in a Tokio task in the same process as the UI. Numbers below are illustrative until we publish measured builds beside shipping artifacts.
Cold-start bar uses the same discipline — measure before launch, publish with releases.
It's free. It's open.
StubHouse is open source under the MIT License. No usage tracking. No telemetry without an explicit opt-in. No "Pro" features held hostage. The full source — desktop app, CLI, and core libraries — is on GitHub. PRs welcome.
$ git clone https://github.com/stubhouse/stubhouse
$ cd stubhouse && cargo build --release
Finished release [optimized] target(s) in 2m 14sReady when the API isn't
Free, open source, and available for macOS, Linux, and Windows.