External API

Odoo is usually extended internally via modules, but many of its features and all of its data are also available from the outside for external analysis or integration with various tools. Part of the Models API is easily available over XML-RPC and accessible from a variety of languages.



If you already have an Odoo server installed, you can just use its parameters


For Odoo Online instances (<domain>.odoo.com), users are created without a local password (as a person you are logged in via the Odoo Online authentication system, not by the instance itself). To use XML-RPC on Odoo Online instances, you will need to set a password on the user account you want to use:

  • Log in your instance with an administrator account

  • Go to Settings ‣ Users ‣ Users

  • Click on the user you want to use for XML-RPC access

  • Click the Change Password button

  • Set a New Password value then click Change Password.

The server url is the instance’s domain (e.g. https://mycompany.odoo.com), the database name is the name of the instance (e.g. mycompany). The username is the configured user’s login as shown by the Change Password screen.

url = <insert server URL>
db = <insert database name>
username = 'admin'
password = <insert password for your admin user (default: admin)>

API Keys

New in version 14.0.

Odoo has support for api keys and (depending on modules or settings) may require these keys to perform webservice operations.

The way to use API Keys in your scripts is to simply replace your password by the key. The login remains in-use. You should store the API Key as carefully as the password as they essentially provide the same access to your user account (although they can not be used to log-in via the interface).

In order to add a key to your account, simply go to your Preferences (or My Profile):


then open the Account Security tab, and click New API Key:


Input a description for the key, this description should be as clear and complete as possible: it is the only way you will have to identify your keys later and know whether you should remove them or keep them around.

Click Generate Key, then copy the key provided. Store this key carefully: it is equivalent to your password, and just like your password the system will not be able to retrieve or show the key again later on. If you lose this key, you will have to create a new one (and probably delete the one you lost).

Once you have keys configured on your account, they will appear above the New API Key button, and you will be able to delete them:


A deleted API key can not be undeleted or re-set. You will have to generate a new key and update all the places where you used the old one.


To make exploration simpler, you can also ask https://demo.odoo.com for a test database:

import xmlrpc.client
info = xmlrpc.client.ServerProxy('https://demo.odoo.com/start').start()
url, db, username, password = \
    info['host'], info['database'], info['user'], info['password']

Logging in

Odoo requires users of the API to be authenticated before they can query most data.

The xmlrpc/2/common endpoint provides meta-calls which don’t require authentication, such as the authentication itself or fetching version information. To verify if the connection information is correct before trying to authenticate, the simplest call is to ask for the server’s version. The authentication itself is done through the authenticate function and returns a user identifier (uid) used in authenticated calls instead of the login.

common = xmlrpc.client.ServerProxy('{}/xmlrpc/2/common'.format(url))
    "server_version": "13.0",
    "server_version_info": [13, 0, 0, "final", 0],
    "server_serie": "13.0",
    "protocol_version": 1,
uid = common.authenticate(db, username, password, {})

Calling methods

The second endpoint is xmlrpc/2/object, is used to call methods of odoo models via the execute_kw RPC function.

Each call to execute_kw takes the following parameters:

  • the database to use, a string

  • the user id (retrieved through authenticate), an integer

  • the user’s password, a string

  • the model name, a string

  • the method name, a string

  • an array/list of parameters passed by position

  • a mapping/dict of parameters to pass by keyword (optional)

For instance to see if we can read the res.partner model we can call check_access_rights with operation passed by position and raise_exception passed by keyword (in order to get a true/false result rather than true/error):

models = xmlrpc.client.ServerProxy('{}/xmlrpc/2/object'.format(url))
models.execute_kw(db, uid, password,
    'res.partner', 'check_access_rights',
    ['read'], {'raise_exception': False})

List records

Records can be listed and filtered via search().

search() takes a mandatory domain filter (possibly empty), and returns the database identifiers of all records matching the filter. To list customer companies for instance:

models.execute_kw(db, uid, password,
    'res.partner', 'search',
    [[['is_company', '=', True]]])
[7, 18, 12, 14, 17, 19, 8, 31, 26, 16, 13, 20, 30, 22, 29, 15, 23, 28, 74]


By default a search will return the ids of all records matching the condition, which may be a huge number. offset and limit parameters are available to only retrieve a subset of all matched records.

models.execute_kw(db, uid, password,
    'res.partner', 'search',
    [[['is_company', '=', True]]],
    {'offset': 10, 'limit': 5})
[13, 20, 30, 22, 29]

Count records

Rather than retrieve a possibly gigantic list of records and count them, search_count() can be used to retrieve only the number of records matching the query. It takes the same domain filter as search() and no other parameter.

models.execute_kw(db, uid, password,
    'res.partner', 'search_count',
    [[['is_company', '=', True]]])


calling search then search_count (or the other way around) may not yield coherent results if other users are using the server: stored data could have changed between the calls

Read records

Record data is accessible via the read() method, which takes a list of ids (as returned by search()) and optionally a list of fields to fetch. By default, it will fetch all the fields the current user can read, which tends to be a huge amount.

ids = models.execute_kw(db, uid, password,
    'res.partner', 'search',
    [[['is_company', '=', True]]],
    {'limit': 1})
[record] = models.execute_kw(db, uid, password,
    'res.partner', 'read', [ids])
# count the number of fields fetched by default

Conversedly, picking only three fields deemed interesting.

models.execute_kw(db, uid, password,
    'res.partner', 'read',
    [ids], {'fields': ['name', 'country_id', 'comment']})
[{"comment": false, "country_id": [21, "Belgium"], "id": 7, "name": "Agrolait"}]


even if the id field is not requested, it is always returned

Listing record fields

fields_get() can be used to inspect a model’s fields and check which ones seem to be of interest.

Because it returns a large amount of meta-information (it is also used by client programs) it should be filtered before printing, the most interesting items for a human user are string (the field’s label), help (a help text if available) and type (to know which values to expect, or to send when updating a record):

    db, uid, password, 'res.partner', 'fields_get',
    [], {'attributes': ['string', 'help', 'type']})
    "ean13": {
        "type": "char",
        "help": "BarCode",
        "string": "EAN13"
    "property_account_position_id": {
        "type": "many2one",
        "help": "The fiscal position will determine taxes and accounts used for the partner.",
        "string": "Fiscal Position"
    "signup_valid": {
        "type": "boolean",
        "help": "",
        "string": "Signup Token is Valid"
    "date_localization": {
        "type": "date",
        "help": "",
        "string": "Geo Localization Date"
    "ref_company_ids": {
        "type": "one2many",
        "help": "",
        "string": "Companies that refers to partner"
    "sale_order_count": {
        "type": "integer",
        "help": "",
        "string": "# of Sales Order"
    "purchase_order_count": {
        "type": "integer",
        "help": "",
        "string": "# of Purchase Order"

Search and read

Because it is a very common task, Odoo provides a search_read() shortcut which as its name suggests is equivalent to a search() followed by a read(), but avoids having to perform two requests and keep ids around.

Its arguments are similar to search()’s, but it can also take a list of fields (like read(), if that list is not provided it will fetch all fields of matched records):

models.execute_kw(db, uid, password,
    'res.partner', 'search_read',
    [[['is_company', '=', True]]],
    {'fields': ['name', 'country_id', 'comment'], 'limit': 5})
        "comment": false,
        "country_id": [ 21, "Belgium" ],
        "id": 7,
        "name": "Agrolait"
        "comment": false,
        "country_id": [ 76, "France" ],
        "id": 18,
        "name": "Axelor"
        "comment": false,
        "country_id": [ 233, "United Kingdom" ],
        "id": 12,
        "name": "Bank Wealthy and sons"
        "comment": false,
        "country_id": [ 105, "India" ],
        "id": 14,
        "name": "Best Designers"
        "comment": false,
        "country_id": [ 76, "France" ],
        "id": 17,
        "name": "Camptocamp"

Create records

Records of a model are created using create(). The method will create a single record and return its database identifier.

create() takes a mapping of fields to values, used to initialize the record. For any field which has a default value and is not set through the mapping argument, the default value will be used.

id = models.execute_kw(db, uid, password, 'res.partner', 'create', [{
    'name': "New Partner",


while most value types are what would be expected (integer for Integer, string for Char or Text),

Update records

Records can be updated using write(), it takes a list of records to update and a mapping of updated fields to values similar to create().

Multiple records can be updated simultaneously, but they will all get the same values for the fields being set. It is not currently possible to perform “computed” updates (where the value being set depends on an existing value of a record).

models.execute_kw(db, uid, password, 'res.partner', 'write', [[id], {
    'name': "Newer partner"
# get record name after having changed it
models.execute_kw(db, uid, password, 'res.partner', 'name_get', [[id]])
[[78, "Newer partner"]]

Delete records

Records can be deleted in bulk by providing their ids to unlink().

models.execute_kw(db, uid, password, 'res.partner', 'unlink', [[id]])
# check if the deleted record is still in the database
models.execute_kw(db, uid, password,
    'res.partner', 'search', [[['id', '=', id]]])

Inspection and introspection

While we previously used fields_get() to query a model and have been using an arbitrary model from the start, Odoo stores most model metadata inside a few meta-models which allow both querying the system and altering models and fields (with some limitations) on the fly over XML-RPC.


Provides information about Odoo models via its various fields


a human-readable description of the model


the name of each model in the system


whether the model was generated in Python code (base) or by creating an ir.model record (manual)


list of the model’s fields through a One2many to ir.model.fields


One2many to the Views defined for the model


One2many relation to the Access Rights set on the model

ir.model can be used to

  • query the system for installed models (as a precondition to operations on the model or to explore the system’s content)

  • get information about a specific model (generally by listing the fields associated with it)

  • create new models dynamically over RPC


  • “custom” model names must start with x_

  • the state must be provided and manual, otherwise the model will not be loaded

  • it is not possible to add new methods to a custom model, only fields

a custom model will initially contain only the “built-in” fields available on all models:

models.execute_kw(db, uid, password, 'ir.model', 'create', [{
    'name': "Custom Model",
    'model': "x_custom_model",
    'state': 'manual',
    db, uid, password, 'x_custom_model', 'fields_get',
    [], {'attributes': ['string', 'help', 'type']})
    "create_uid": {
        "type": "many2one",
        "string": "Created by"
    "create_date": {
        "type": "datetime",
        "string": "Created on"
    "__last_update": {
        "type": "datetime",
        "string": "Last Modified on"
    "write_uid": {
        "type": "many2one",
        "string": "Last Updated by"
    "write_date": {
        "type": "datetime",
        "string": "Last Updated on"
    "display_name": {
        "type": "char",
        "string": "Display Name"
    "id": {
        "type": "integer",
        "string": "Id"


Provides information about the fields of Odoo models and allows adding custom fields without using Python code


Many2one to ir.model to which the field belongs


the field’s technical name (used in read or write)


the field’s user-readable label (e.g. string in fields_get)


the type of field to create


whether the field was created via Python code (base) or via ir.model.fields (manual)


enables the corresponding flag on the field


field-level access control, a Many2many to res.groups


type-specific properties and customizations, see the fields documentation for details

Like custom models, only new fields created with state="manual" are activated as actual fields on the model.


computed fields can not be added via ir.model.fields, some field meta-information (defaults, onchange) can not be set either

id = models.execute_kw(db, uid, password, 'ir.model', 'create', [{
    'name': "Custom Model",
    'model': "x_custom",
    'state': 'manual',
    db, uid, password,
    'ir.model.fields', 'create', [{
        'model_id': id,
        'name': 'x_name',
        'ttype': 'char',
        'state': 'manual',
        'required': True,
record_id = models.execute_kw(
    db, uid, password,
    'x_custom', 'create', [{
        'x_name': "test record",
models.execute_kw(db, uid, password, 'x_custom', 'read', [[record_id]])
        "create_uid": [1, "Administrator"],
        "x_name": "test record",
        "__last_update": "2014-11-12 16:32:13",
        "write_uid": [1, "Administrator"],
        "write_date": "2014-11-12 16:32:13",
        "create_date": "2014-11-12 16:32:13",
        "id": 1,
        "display_name": "test record"
0 总浏览
0 成员的观点
0 公众浏览
0 喜爱
0 不喜欢
0 注释

登录 分享这个 webpage 通过电子邮件。