Rich Birks
Published on 22 May 2025
ScriptRunner Connect
Accurate, scalable, future-proof: The ScriptRunner Connect AI assistant
Whether you're a seasoned ScriptRunner pro or just getting started, this is the time-saving tool you need to improve the scripting process and master even the most complex integrations.
Let's be honest: you're busy, and your time is precious. Our job is to help you do your job, by simplifying, speeding up, and smoothing out the tasks that take too much of your time and mental bandwidth away from the important stuff. That's the principle at the core of the whole ScriptRunner suite. The latest progress in pursuit of that lofty goal is getting your app integrations up and running more easily and quickly on ScriptRunner Connect—with a built-in AI assistant.
We know, we know, everything has an AI assistant now, but this is one of the good ones. Promise. The ScriptRunner Connect assistant isn't trying to reinvent the wheel. It's been designed specifically to help with the work you're already on the platform to do: building integrations through the power of scripts, APIs, and webhooks; as well as learning the product's full capabilities and finding relevant out-of-the-box use case templates.
The ScriptRunner Connect assistant is specifically trained with the API surface of the ScriptRunner Connect runtime and its Managed APIs (MAPIs), so it provides a similar experience to other LLM-based chatbots, but with the accuracy of a product-specific knowledge base. If you're not familiar with MAPIs, they're a set of handy abstractions that simplify the process of making API requests to third-party apps. The AI assistant also pulls in public domain knowledge to fill in any gaps, troubleshoot errors, explain code, and even suggest improvements for existing code.
In other words, the assistant makes it easier than ever to create ready-to-run scripts and advance towards that fabled technological utopia where all your tools are securely, seamlessly connected.
But we're getting ahead of ourselves. What makes this better than any of the dozens of standalone LLM chatbots scattered across the technoverse already? That's a valid question, and it comes down to three things: accuracy, scalability, and future-proofing.
Accurate
In building an AI assistant for ScriptRunner Connect, accuracy was top of the list. You need to know that the AI not only fully understands what you need from it in this context, but also that you can rely on its output to actually be useful.
Since each API method needs its own documentation—and Jira Cloud alone has over 350 methods—the assistant uses documentation automatically generated from our Managed API packages, ensuring every supported method is included.
However, the assistant still needs to generate accurate code in cases where documentation is missing. Built-in evaluation logic assigns confidence scores to all retrieved documentation, and for anything with a low confidence score, the assistant will reference API documentation from the public domain while using ScriptRunner's own fetch APIs.
Put simply, rather than relying on potentially irrelevant documentation, the assistant will default to using the generic fetch method, and you'll always receive actionable guidance—even if the retrieval logic slips up a little, or your particular API isn't natively supported in ScriptRunner Connect. And for completely unsupported apps where no documentation can be found? The assistant will leverage a Generic API handler, so you can still make API calls using a standardised approach.
Having such a robust evaluation and fallback mechanism dramatically reduces hallucinations, making this AI assistant genuinely helpful.
Scalable
Next up: scalability. ScriptRunner Connect already has MAPIs for all of your popular tools across ITSM, DevOps and work management practice areas, such as Azure DevOps, Bitbucket, Confluence, GitHub, GitLab, Jira, JSM, NetSuite, Salesforce, ServiceNow, Trello, and Zendesk. In fact, the assistant is trained on and supports over 20 different third-party applications or MAPIs as of right now, and beyond that, it can still confidently fall back on public domain knowledge and point you in the right direction. With new tech being regularly added to your toolkit, the AI needs to keep up—otherwise it's just another icon gathering dust and taking up space in your sidebar.
So, instead of having hardcoded support for specific APIs just fed into context, your ScriptRunner Connect AI assistant makes use of a flexible retrieval system with a vector database. To give you a little peek behind the curtain, each MAPI method is stored under a dedicated namespace (e.g., jira-cloud). When you submit a prompt, the system breaks it down into logical API requests (AKA retrieval-augmented generation, or 'RAG'), assigns a namespace, and retrieves the best matching documentation. This means the assistant can efficiently retrieve relevant documentation, even as new apps are integrated, giving you a whole system that is scalable.
As ScriptRunner Connect expands, your friendly neighbourhood AI assistant will seamlessly support new apps without requiring major overhauls. We're also adding more all the time—just ask if you have something in mind!
Future-proof
Finally, there's future-proofing. The scalability considerations above are one part of the equation, but even more so, we know that the AI models at the core of assistants like this one are only going to continue improving. And you want the latest and greatest, right?
The way this assistant is built, there's one model that handles the prompts, and another that handles middleware queries to optimise process time and optimise your requests. Middleware queries include defining the type of request, whether it be for code generation or querying the knowledge base, and breaking down complex queries through query expansion. Handling those intermediary steps separately means that the core chatbot logic remains predictable, so as models improve, it's simple to update the LLM version.
Your ScriptRunner Connect chatbot can (and will!) get faster, more powerful, and even more reliable without any action needed from you. When we update the LLM version, your code companion's capabilities get upgraded, with no need to change the core retrieval and evaluation logic that makes it so good—so you'll always have the best possible experience.
Haven't tried it yet? Get stuck in!
Describe what you need in plain English, and let your new AI companion handle the rest. Get the flexibility and precision of advanced coding capabilities with the simplicity and speed of an AI-assisted environment.
Give it a try today
ScriptRunner Connect brings efficiency and alignment to your teams by building every app integration securely within a central platform.