MuleSoft AI Chain is a Product now

?? ??????????????, ???????????????? ???? ?????????? ?????? ???????? ???????????????????? ??????????????????!?? ??

Yesterday, MuleSoft AI Chain Connector has been officially released on your MuleSoft Anypoint Exchange. This is a big milestone for the hard work and relentless contributions of the whole team!

The Beginning

This started in mid ?????? 2024, with the ambition to open up MuleSoft's Lifecycle Capabilities to AI Agents & Workflow! I discovered a Java Library called LangChain4J (if you don't know it, check it out). Similar to LangChain, LlamaIndex and other AI orchestration frameworks, but for Java. I implemented the first connector for ?????????????????? ???? (we later renamed it to ???????????????? ???? ??????????) and immediately told Mihael Bosnjak about it. Miki was ??????-???? and we initiated ?????? ?????????????? (???????????????????? ?????????????????? ??????????????) and started building mac-project.ai.

Later on, Ajay Rana joined to help us shape this project, along with Jeroen Reizevoort for the strategic influence and advisory and Vibhor Sharma for helping us with the code! We launched the Project on the 1st of June 2024 as an open-source: https://lnkd.in/eyy5QJvM

We presented the following vision for the MuleSoft AI Chain Project, which today is a little different.

Ajay Kumar our own mister AI became our first advisor and ambassador for the project right after the release and he invited me to World Tour UK, where I met Kranthi Pullagurla. Kranthi heard about this project and asked me to look into Amazon Bedrock support. This initiated a parallel stream to co-develop Amazon Bedrock Connector for MuleSoft, and making the Project more compelling in the AI space. We presented the vision to our Field CTO Prashant Choudhary , who guided us into the right direction to make it more compelling and join the advisory board with Ajay!

Visibility and Traction

At that time we did not know, it will be officially certified by MuleSoft and I think it wasn't ready. ? Damian Scherrer joined the team to help us improve the connectors.

With the addition of Burak Tas and Jonathan Chen in June, we started speaking more about it. Mid June, we've started discussions with Product & Engineering Teams to productize the Connector led by Ryan Stastny. Dipesh Dutta and Arpit Gupta joined the team! They started to clean the code and prepare the connector for certification. Rohan Vettiankal started to shape the official GTM assets and helped us become more compliance.

Early July Viktoriya Kotik joined and she took us to the next level, by promoting it internally to the right stakeholders and setting the stage for external audiences! The interest internally as well as externally in MuleChain started to increase! Project related discussions started and Jaime Martín Talavera joined, along with Ahmed-Jibril Ait ouaret ??, Laura de Paz Carbajo. With the help of Antonio Galdo, Lukas Auge and Dan Porat we were able to present the solution in early stages, and started talking about enterprise use cases with customers in Spain, Switzerland and Israel. Together with Laura and Jonathan, we built our first Documentation Agent which later was transformed into an RFP agent! Thats one of the best internal case studies, I have experienced! At the same time Ahmed-Jibril Ait ouaret ?? improved the Einstein AI Connector and MAC Whisperer! Beginning of August Chuck Hirstius joined and immediately Chuck took the project to different heights by enabling agent characteristics and building MAX, the first accelerator of the project! Chiew Kwooi Lee joined few days later in August and we had established a great foundation now with Chiew to provide Example Mule Apps for the community.

Refinement of Vision

At this time, the team had corrected the initial vision to fully go into the AI Landscape and concentrating on what matters.

We could feel more traction and that the project has took off. With Yogesh Mudaliar and Alick Wong joining we accelerated the development and WebCrawler and Vectors became reality! As the project was making a lot of noise, Sally H'Doubler joined to help us shape it internally towards a program and help us get sponsorship of Jennifer McLaughlin and Wolfgang Kelz! We were happy to have the leadership support for us with a much stronger backbone, then without!

The Take Off

Many contribution happened in August with Akshata Sawant promoting us through the developer channels, in September Ettore Giallaurito brought use cases for advanced and agentic RAG, and we improved the connectors. When we started to build advanced RAG, Tommaso Bolis got involved and implemented RAG use cases with MuleSoft. Right after in October Tommaso joined the Project and started to make the Vectors connector enterprise ready! The contributions continued, with James Losack building some great use cases on agentic workflows for data ingestion, the project had started to deliver industry use cases. Prashanth Kurimella delivered an amazing demo on how MuleSoft is extending Agentforce with the MAC Project and immediately after joined the team.

We've also onboarded many community contributors with Joost Damen , Jovana Vujani? , Beauty Mishra , Isha Sharma , Tanupam Saha , Dhairya Taneja , @Fatima Maniyar who fueled the interest in the MAC Project by their valuable contributions with use cases and demos!

Recently the teams from LPDG with Tobias Stein , Dr. Christian Berker and Julian G. joined, along the Exavlu team led by Brojeswar Roy with Sandeep Shenoy, Himasree Anigondi, Pratik Biswas, Saptarshi Chakraborty, Rohit Nandy, @Arun Saha, Rajendra (Raj) Kelkar, Rajesh Gowrineni and Naveen Korutla.

Ryan Hoegg joined last week and is currently supporting the open source project on simplifying the installation process. And we are growing in the MuleSoft Community.

With Carlos Casares , Mizuho Hoshino (Yamamoto) , Queena(Qianwen) X. , Michael Tietze, @Anand Shankar Uday Shankar, @Ken Chang, and @Raymond Arnado, the team has grown and continuing to contribute at all areas internally, and externally.

This post dedicated to all contributors! I want to congratulate everyone who contributed to the MAC Project for all the hardwork, contribution and relentless support to make it become a official support! THANK YOU, without the contributions, we would not be here.

You can find the officially certified and productised connector on your Anypoint Exchange.

You have noticed that the icon changed, and thats going to happen to all connectors on certification. Still we will keep the spirit and legacy fully documented on mac-project.ai.

Whats Next

The question remains, are we done? Where are we going from here? The MuleSoft AI Chain project will remain open source and we will continue to release features and improvements on the open source repositories for all connectors. The features will be certified and at some point added to the officially supported versions as well. This means, we are just getting started and there is a lot more to come for the project!

Some useful resources

• Project website: https://mac-project.ai/
• Meet the Team: https://mac-project.ai/about
• GitHub Org: https://github.com/MuleSoft-AI-Chain-Project
• YouTube Channel: https://www.youtube.com/@MuleSoft-MAC-Project
• Official Connector on Exchange: https://anypoint.mulesoft.com/exchange/com.mulesoft.connectors/mule4-aichain-connector/minor/1.0/
• Official Connector Documentation: https://docs.mulesoft.com/mulesoft-ai-chain-connector/latest/