Introducing Livechat for CSML

Bots are great, but sometimes, there are things where you just need a little extra help from a real human. That's when you want to stop talking to a bot and reach a real human instead.

Introducing Livechat for CSML Studio. You can now switch to livechat with a human agent without ever leaving the chatbot conversation. Using Livechat, you can now better support your users and propose a better experience all around by never ever leaving a user in the dark.

Livechat relies on the best providers of customer support software. The launch partner of this feature is Gorgias, a customer support platform dedicated to e-commerce and tightly integrated with Shopify.

Livechat is available on most channels: Facebook Messenger, Workplace Chat, Microsoft Teams, Slack, Whatsapp and of course our own webapp and chatbox plugins.

Try the Gorgias integration with Livechat for CSML Studio today ! Documentation available here:

New feature: Bring your own NLP!

Great news! CSML Studio now supports multiple NLP providers natively! You can now bring your own NLP into your CSML chatbot with just a few clicks.


Previously, to support NLP, you had to install an app or a custom function and use it as an event preprocessor. Now, you can simply go to Bot Settings > NLP, select your NLP provider of choice, and all your chat events will now automatically be preprocessed with your configured NLP provider.

For the launch of this fantastic feature, CSML Studio supports the following providers:

  • Dialogflow
  • SAP Conversational AI (formerly
  • Amazon Lex
  • Microsoft LUIS
  • Custom HTTP Webhook

Many more providers are underway. Visit the documentation to learn more about this great and much awaited feature!

CSML Open Source V1

The CSML Conversational Engine is now officially launched in v1!

We are proud to launch the v1 of CSML Conversational Engine! After months of production use and millions of messages exchanged with hundreds of thousands of users, we are incredibly proud to have reached a first major milestone on our journey.

In these past few months, we have learned so much about chatbots and conversational technologies, and are more convinced than ever that using a high-level language such as CSML is the only way to develop high-performance, full-featured, maintainable chatbots. Our public beta launch post is even more relevant now, with so many businesses and parts of our lives going online, and we need powerful tools to help with that.

With this release, you can now install all the tooling that makes CSML so great directly on you own machines, using free, open-source solutions available to everybody. We want to make rich chatbot building more accessible to anyone.

This is only the beginning of the journey! Stay tuned for more to come.

In the v1 release, you get, for free and forever:

  • CSML Interpreter
  • Conversational Engine
  • Context handling
  • Short/Long-term memory handling
  • Easy-to-run docker image
  • MongoDB connector
  • Custom nodejs or generic AWS Lambda Fn runtimes
  • Lots of new bots ready to use

Go to to get started with CSML Conversation Engine!


In a chatbot, there are some situations where you want to schedule an action for later. A simple example: if the user left some items in their shopping cart, remind them about it in a few hours to drive your sales.

We just released an important new feature with the Scheduler. You can now schedule actions to be performed in 1 min to 1 month. There are two ways to use the scheduler:


At the scheduled time, POST a request to the given HTTP endpoint, with an optional body.

Fn("utils/scheduler", action="webhook", schedule_time="2020-03-12 12:33:00", body={"name": "Arthur"})


At the scheduled time, send a broadcast to the current user with the given flow and optional metadata. NB: the broadcast scheduler only works on channels that support broadcasts (messenger, workchat, msteams).

Fn("utils/scheduler", action="broadcast", delay="3 days", flow_id="myFlow", metadata={"message_tag":"ACCOUNT_UPDATE"})

Obviously, you can cancel a scheduled item at any time by giving it a schedule_id and canceling this action later, for example if the user goes to actually buy the items in their shopping cart:

Fn("utils/scheduler", action="cancel", schedule_id="cart_reminder")

New integrations: Airtable, Gorgias, Huggingface NLP, Picture Frame

A bunch of new integrations this month!


You can now easily connect your CSML chatbot to an Airtable database. Airtable is a nice, user-friendly mix between Excel and a regular database. We published an article here to get you started with Airtable and CSML!


We are preparing some very nice updates with customer service solutions, and we are first adding a Gorgias integration to make it possible to create tickets in Gorgias directly from the confort of a CSML chatbot. If this sounds interesting, stay tuned ;-)


So, Huggingface is currently the hottest Natural Language Processing startup around. You should definitely check out their new online models feature! And they are doing a LOT of open-source work, which we obviously loved.

So we hooked a sentiment analysis and question answering model from Huggingface in a CSML app. Try it out now!

Picture Frame

Let users add a custom frame on top of their profile picture with this nice app. Simply have a png ready to overlay on top of a given picture, and voilà, your users can now download their own profile picture!


We even created a free chatbot for Workplace with some custom Summer-themed frames that you can easily get started with:

So, what do you think? If you have any cool idea for a CSML app, get in touch at !

Messenger Broadcasts

You can now use CSML Studio to send messages proactively to your users. Broadcasts is a very powerful feature: it's a fantastic way to reengage with your users, and get them to interact again with your bot.

Please note that the standard limitations apply: while you can send practically any message to any user within 24 hours of their last interaction with the chatbot, messages sent outside of this window must comply with strict rules as explained here.

Broadcasts can be sent either from the API (great for engagement automation!) or from CSML Studio as shown below:


Let us know what you plan to build with Messenger Broadcasts!

New integrations: Clickup, Netflix, Jitsi, DynamoDB, QR Code

New integrations have been added to the apps directory this week:

  • Clickup lets you access all your team's tasks from your chatbot

Capture d'écran 2020-06-13 18.58.08.png

  • Search for movies in Netflix catalog

Capture d'écran 2020-06-13 18.58.43.png

  • Jitsi lets you easily create secure video conferences links without any registration, for free

Capture d'écran 2020-06-13 18.59.12.png

  • Generate QR Codes directly inside the chatbot conversation

Capture d'écran 2020-06-13 19.00.37.png

  • Manage data securely in your own Amazon DynamoDB table

Capture d'écran 2020-06-13 18.58.44.png

UX improvements

We recently released a lof of UX improvements all across CSML Studio. In no particular order:

  • New sidebar bot selector, allowing to search for your bot and scroll for users with many bots
  • New channel creation workflow, with fewer clicks until channel is created
  • Improved channel configuration page, splitting channel configuration and API access information
  • Improved bot settings, splitting bot and preprocessing configuration for an easier experience
  • Merged functions and apps directory to the same panel for easier navigation between both
  • New app tag on functions installed from the apps directory

Let us know what you think of all these improvements!

Google Signin

You can now use Google to sign into your CSML studio account. Much easier!

Capture d'écran 2020-06-13 19.09.50.png

Broadcasts API

You can now use the API to send broadcasts to your users on your channel of choice. Previously, you had to manually login to the platform and add your targets in the channel's settings page, as shown below.

Capture d'écran 2020-06-01 15.31.08.png

This is very easy to use, but does not allow for automatisation. With this update you can now perform HTTP requests to create broadcasts programmatically.

Please refer to the documentation to learn more about the Broadcasts API.

Of course, you can still send broadcasts from the interface!

Each channel has its set of features and limitations. For example Messenger and MS Teams only allow sending to users that have already interacted with your bot before.