stubfetch

Deterministic, in-process fake HTTP APIs for testing agents and tools. Swap live network calls for canned GitHub, Stripe, OpenAI, Anthropic, S3, and Slack responses — while recording traffic, running eval scenarios, and injecting chaos.

Install

npm install stubfetch

Requires Node.js 18+. No runtime dependencies.

Quick start

import { GhostEnv, github, exportRecordingJSON } from "stubfetch";

const env = new GhostEnv({
  seed: 42,
  providers: [github({ issues: [{ repo: "acme/api", title: "Bug #1" }] })],
});

const res = await env.fetch("https://api.github.com/repos/acme/api/issues");
const issues = await res.json();
console.log(issues[0].title); // "Bug #1"

// Inspect call history
console.log(env.wasCalled("github")); // true
console.log(exportRecordingJSON(env.calls()));

Presets

Preset	Import	URLs handled
GitHub	`github()`	`api.github.com` issues, repos, PRs
Stripe	`stripe()`	`api.stripe.com` customers, charges
OpenAI	`openai()`	`api.openai.com` chat completions
Anthropic	`anthropic()`	`api.anthropic.com` messages
S3	`s3()`	`s3.amazonaws.com` GET, list, put
Slack	`slack()`	`slack.com/api` auth, messages
Postgres	`db()`	In-process `db().query()`

Chaos injection

const env = new GhostEnv({
  seed: 0,
  providers: [github()],
  chaos: { minLatencyMs: 50, failureRate: 0.1 },
});

Guides

Doc	Contents
Overview	Product summary and sidebar
Getting started	Providers, fetch, recording (Python: `fetch` → `(status, body)`)
Use cases	Stubs, fixtures, evals
Presets	URLs, config shapes, matching rules
API reference	`GhostEnv`, `Provider`, eval, replay, exports
Testing & chaos	`runEval` / `run_eval`, chaos, failure recording
Python package	Node vs Python differences