How to write a connector


New! Want an interactive tutorial on how to create a connector? Try this one:

A connector is a small script that allows to import data from an external website. Each connector is an independent application, managed by the Cozy Collect application.

To protect your data, each connector runs inside a sandbox where all their interactions with your data are under control.

From a technical point of view, connectors are Node.js applications executed inside a container. They communicate with the server using its API, like client side apps, and get an auth token every time they start. They need to register with a manifest, and ask permissions to the user.

To ease the development of a connector, an npm package, named cozy-connector-libs provides a lot of shared libraries. But you may need some other npm packages to help you running your connector:

  • cheerio to manipulate the DOM on remote web pages ;
  • moment to manage dates ;
  • request for fetching remote URLs ;

When the application is started, it also gets some data through environment variables:

  • COZY_CREDENTIALS : the auth token used by cozy-client-js to communicate with the server ;
  • COZY_URL : the API entry point ;
  • COZY_FIELD : the settings specific to each connector, for example the path of folder where the user wants to store the remote files ;

But the base connector (require(‘cozy-konnector-libs’).baseKonnector) in cozy-konnector-libs handles these for you.

The application can access a temporary file system, deleted at the end of its execution. Its logs (standard and error output) are kept by the server.

From the server point of view, each connector is a job run through a trigger.

Let’s create our first connector

The easiest way to create a new connector is to use our template:

git clone
cd cozy-konnector-template
yarn # or npm install

note: the Cozy Team uses yarn, but if you prefer npm, just keep using it, everything should just work.

Then, write your code into konnectors.js and build the application, running yarn build or npm run build.

Collector structure

Basically, a connector is just an object passed to baseKonnector.createNew:

  • fetchOperations: array of methods that will be called sequentially to fetch the data;

To create the connector, just call baseKonnector.createNew() from the Cozy Collector lib, with an object describing the connector:

const {baseKonnector, filterExisting, saveDataAndFile, models} = require("cozy-konnector-libs");
const Kitten = models.baseModel.createNew({
  name: "kitten"

module.exports = baseKonnector.createNew({
  name: "kitten",
  models: [],
  fetchOperations: [

Fetch operations

Every time the connector is run, it will call every method from the fetchOperations array. Use this methods to log into the remote site, fetch data and save it.

Each function must use the same signature: functionName(fields, bills, data, next) where:

  • fields are the values of the optional configuration fields;
  • entries is an object to pass data from one function to the next one;
  • data allows to pass raw data from one function to the next one;
  • next is a function to call to execute next function. One may pass an error as first argument. Don’t forget to call it at the end of every step.

A basic connector workflow involves: - getting data and storing them into entries.fetched. You can get the data by calling an API, scraping the remote website…; - filtering data to remove the ones already present inside the database. The filtered data will be put into entries.filtered; - save the filtered data into the database.

Many operations are common to most of the connectors, so we created some common functions you can import from the shared library. Most of the time, you’ll just have to take care of fetching data, store them into entries.fetched then use the common methods to filter and save them.

We’ll have a deeper look at this methods below.

Error handling

If your connector hit some issue fetching or saving the data, it can return an error code by passing it to the next method. Some error code are defined inside the Cozy Collect application and will display an explicit error to the user:

  • LOGIN_FAILED: the konnector could not login;
  • NOT_EXISTING_DIRECTORY: the folder specified as folder_to_save does not exist (checked by base_konnector);
  • UNKNOWN_ERROR: there was an unexpected error, please take a look at the logs to know what appened;

Konnector lib

The Cozy Konnector Lib provide some useful methods for common tasks:

  • baseKonnector.createNew(): create the connector and fetch data;
  • cozyClient gives an instance of cozy-client-js already initialized according to COZY_URL, and COZY_CREDENTIALS
  • fetcher is the internal class that run fetching operations in sequence, calling the functions with the right parameters;
  • log(type, message) allows to log messages;
  • manifest extracts informations from the manifest (mainly used internaly at the moment);
  • naming is a method allowing to build file names according to parameters;
  • filterExisting to filter data;
  • linkBankOperation to link a bill to a bank operation;
  • saveDataAndFile save the data;
  • updateOrCreate create or update documents inside database

There are also some models available in require(‘cozy-konnector-libs’).models :

  • baseModel : which is a model from which your model can extend using it’s createNew method. It offers the all, create, and updateAttributes methods already implemented.
  • bill : Offers an already implemented model allowing to manipulate ‘io.cozy.bills’ doctype. It is used by many connectors.
  • file : a helper to manipulate files
  • folder : a helper to manipulate folders
  • bankOperation : a helper to manipulate bank operations

Common methods


If you want to access cozy-client-js directly, this method gives you directly an instance of it, initialized according to COZY_URL and COZY_CREDENTIALS environment variable. You can refer to the cozy-client-js documentation for more information.

const {clientClient} = require('cozy-konnector-libs')'my.doctype', ['_id'])

filterExisting(log, model, suffix, vendor)

This method returns a fetch function that filter data fetched from the remote site to only keep the ones that don’t exist in database. The fetched data are expected to be in the entries.fetched array. The resulting array will be put into entries.filtered


  • log: unused (kept for retro-compatibility);
  • model: the model;
  • suffix: unused (kept for retro-compatibility);
  • vendor: if a vendor parameter is given, entry should be of given vendor to be added to the hash (useful for bills).
konnector.fetchOperations = [ (…), customFilterExisting, (…) ];

function customFilterExisting(requiredFields, entries, data, next) {
  filterExisting(myKonnector.logger, Bill) (requiredFields, entries, data, next);

saveDataAndFile(logger, model, options, tags)

This method returns a fetch function that creates an object in database for each item in entries.filtered array. If item has a pdfurl attribute, the remote file will be downloaded and stored on the filesystem. pdfUrl can point to any file, not necessarily a PDF file. The name comes from legacy code and has not been updated.


  • log: unused (kept for retro-compatibility);
  • model: the model;
  • options:
  • tags: array of tags to apply to created files.

updateOrCreate(logger, model, filter, options)`

This method return a fetch function that creates or updates an object in database for each item in the entries[model.displayName] array. The filter parameters specifies the fields used to find the document inside the database.


  • log: unused (kept for retro-compatibility);
  • model: the model;
  • filters: an array of fields names;
  • tags: array of tags to apply to created files.


This method returns a fetch function that will try to link a bill to a bank operation. For each data item from entries.fetched, it will look for an operation that could match this entry. Once found, it attaches a binary to the bank operation. It’s the same binary that is attached to the corresponding file object.

The criteria to find a matching operation are:

  • Operation label should contain one of the identifiers given in parameter;
  • The date should be between (bill date - dateDelta) and (bill date + dateDelta). Where dateDelta is given as a parameter and is in days;
  • The amount should be between (bill amount - amountDelta) and (bill amount + amountDelta). Where amountDelta is given as a parameter.


You should pass parameters as an object whose keys are:

  • log: unused (kept for retro-compatibility);
  • model: a model object to check for.
  • identifier: a string or an array of strings to look for in the operation label (case insensitive: the layer will automatically set it to lowercase).
  • dateDelta: the number of days allowed between the bank operation date and the bill date (15 by default).
  • amountDelta: the difference between the bank operation amount and the bill amount (useful when the currency is not the same) (0 by default).
  • isRefund: boolean telling if the operation is a refund or not. By default, it is false. Allows to match only operations with positive amount if set to true.

Common data models

The library includes the most used data model, so you can just require them if you need them:

  • bankOperation
  • bill
  • file
  • folder

The library also provide a baseModel class to create your own data model. Each model inherits from the following methods:

  • all(callback) fetch all documents;
  • create(entry, callback) creates a new document;
  • updateAttributes(id, changes, callback) update the attributes of a document.
const { models: { baseModel } } = require(’cozy-konnector-libs’)

module.exports = baseModel.createNew({
    displayName: "myModel",
    name: "me.cozy.mymodel"

The manifest

Each connector is described by a Manifest. This is a JSON file named manifest.konnector at the root of your code folder. It should include the following information:

  • name: …
  • slug: the internal name of the application;
  • type: for now, the only allowed value is node. In the future, we may support other types of connectors;
  • version:
  • source: git URL of the source code repository;
  • fields: @TODO
  • locales: @TODO
  • permissions: an object describing the permissions the connector requires;
  • developer: who are you?
  • name:
  • url:


TODO See documentation of the manifest of an application

The connector parameters are stored in io.cozy.accounts documents, so each connector should get access to this doctype.


How do I scrap a website

You will require the request and cheerio npm packages:

yarn add cheerio request # or npm install --save cheerio request

Here’s a sample code that will fetch the login page to get the value of the anti-CSRF token, submit the login form, browse to the bills page and fetch a bill:

function fetchBill(requiredFields, entries, data, next) {
  'use strict';
  // Create a request instance that keep cookies between requests and follow redirects
  let request = require('request').defaults({jar: true, followRedirect: true, followAllRedirects: true}),
      cheerio = require('cheerio'),
      moment  = require('moment');
  // Get the login page to get the CSRF token
  request("https://login.remote.web", function (err, res, html) {
    if (err) {
      return next(err.message);
    // Post the form
    let $ = cheerio.load(html),
        form = {
          form: {
            login: requiredFields.login,
            password: requiredFields.password,
            csrf_token: $('[name="csrf_token"]').val(),
        };'https://login.remote.web', form, function (err, res, html) {
      if (err) {
        return next(err.message);
      request("https://admin.remote.web/bills", function (err, res, html) {
        if (err) {
          return next(err.message);
        entries.fetched = [{date: moment($("bill_date")), value: $("#bill_value")}];

The whole connector will be as simple as:

const {baseKonnector, filterExisting, saveDataAndFile, models} = require("cozy-konnector-libs"),
      MyBills = models.baseModel.createNew({
          name: "me.mycozy.mybill"

function fetchBill(requiredFields, entries, data, next) {

module.exports = baseKonnector.createNew({
  name: 'me.mycozy.mybill',
  fetchOperations: [
    filterExisting(null, MyBills),
    saveDataAndFile(null, MyBills, {})


Running in standalone mode

To ease the development, you don’t need a running Cozy server to test your code. We provide a standalone mode, that mocks the server. This mode uses a configuration file to define the environment variables that the server will send to your application in production. So, start by copying data/env_fields.json.template to data/env_fields.json and set the parameters your application requires. Then start it with yarn standalone or npm run standalone.

In standalone mode, saving a file will put it into the data folder at the root of your repository. If you need to query the database, put your mock data into data/fixture.json. Also, fetched data will just be outputted to the console instead of being sent to the database.