🐵 A framework for fully automatic monkey testing/fuzz testing of GraphQL APIs.
GQLM explores GraphQL schemas with almost zero human input, discovering bugs, edge cases, and security issues.
In an existing Node.js project, run:
npm install gqlm
Or, if you want to set up a new project just for running tests on an API, run these in a new directory:
npm init -y
npm install gqlm
Create a file test.gqlm.mjs
with the following contents:
import { GQLM } from 'gqlm';
GQLM.run({
url: 'https://my.graphql.api/graphql',
count: 5,
});
Then, run node test.gqlm.mjs
.
This will start automatic, explorative testing of your API with up to 5 requests.
- Test results with requests and responses are written to
__gqlm__/<n>.json
. - Collected memory is written to
__gqlm__/memory.json
. - Schema coverage is written to
__gqlm__/coverage.json
. - Inspect these files to determine fitness of the tested GraphQL API.
- A good workflow is to follow up by creating integration/snapshot tests for conservation of behavior.
Example result:
export const operation = `{
search(q: "Blades of the Darkmoon") {
__typename
... on Company {
contracts {
customer {
__typename
... on Company {
employees {
firstname
lastname
birthdate
__typename
}
}
}
}
}
}
}`;
export const result = {
status: 200,
responseTime: 3,
failed: false,
data: {
search: [
{
__typename: 'Company',
contracts: [
{
customer: {
__typename: 'Company',
employees: [
{
firstname: 'Dark Sun Gwyndolin',
lastname: '?',
birthdate: '2011-09-23',
__typename: 'Person',
},
{
firstname: 'Darkmoon Knightess',
lastname: '?',
birthdate: '2011-09-24',
__typename: 'Person',
},
],
},
},
],
},
],
},
};
For secured GraphQL APIs, you are free to authenticate in any way and pass headers to the GQLM options. For example:
import { GQLM } from 'gqlm';
async function run() {
const authResponse = await fetch('https://my.oauth.api/token', {
headers: {
'content-type': 'application/json',
},
body: JSON.stringify({
grant_type: 'password',
username: 'foo',
password: 'bar',
}),
});
const { access_token } = await authResponse.json();
await GQLM.run({
url: 'https://my.graphql.api/graphql',
requestInit: {
headers: {
authorization: `Bearer ${access_token}`,
},
},
count: 5,
});
}
GQLM accepts the following options:
url: string
- URL of GraphQL API to testcount: number
- Maximum number of requests to runrequestInit?: RequestInit
- Additional parameters for fetch(), e.g. headersseed?: string
- Seed for randomization; if set, GQLM will behave deterministicallyinput?: Record<string, unknown>
- Map of initial data for generating inputs.exit?: boolean
- Should GQLM exit on the first failed request?isFailure?: (result: ExecutionResult) => boolean
- Callback to determine whether a GraphQL result is considered a failure.outDir?: string
- Output directory; defaults to__gqlm__
.
Expanding the pattern for authentication above, you can design arbitrary environments around GQLM, e.g.
- starting a local server to test,
- seeding data,
- running with multiple users of different access levels,
- use GQLM for load testing,
- etc.
Additionally, you can extend the GQLM class and override some behavior (see source code).
Currently, only GraphQL queries are supported. Mutations need some discovery on how to test them safely and make the results useful. PRs welcome!
Test files can also be written in TypeScript (.ts
extension).
Just install and use ts-node
instead of node
.