What's New
New assets and notable releases, newest first. Subscribe with the RSS feed to hear about them as they land.
Studio AI 0.5.11: Diagnose modeA new Diagnose mode finds why something goes wrong, with evidence: it reproduces the problem against your running app, reads the console and the Mule XML at the line the error names, checks the cause, and offers the fixes for Agent mode. Calls to the running app now include what its console printed.
- Studio AI 0.5.9: call the running app and read MuleSoft docs
The AI can list your app's HTTP endpoints, call the app you are running in Studio on localhost, and compare the response with what you expect. It reads MuleSoft documentation pages to check a connector's parameters instead of answering from memory.
- Studio AI 0.5.8: Feedback in the view
Feedback in the MuleSoftForge Studio AI header opens a bug or feedback form. Open on GitHub fills a public issue on the plugin's repository for you to create. Secrets are masked; leave out customer data.
- Studio AI 0.5.7: debug with AI, and DataWeave checked with MuleSoft's CLI
When Studio's Mule debugger stops, Ask about this event attaches the live payload, attributes, variables and error, and the AI tests its fix against that event before proposing it. With MuleSoft's DataWeave CLI, proposed DataWeave is compiled and scripts are run on sample input. Accept now works for proposals that create files or folders.
- Studio AI 0.5.6: cancel from the Progress view, Studio themes, clearer errors
Each answer runs as a Studio job you can cancel from the chat or the Progress view. The view follows Studio's light and dark themes, with its colours under Colors and Fonts. Accept all writes every file in one step, and network errors say what to fix.
- Studio AI 0.5.4: standing instructions from AGENTS.md and Preferences
Team conventions in the project's AGENTS.md and your own instructions in Preferences go with every message, with secrets masked. They shape style and Mule conventions; the rules that keep secrets masked and changes reviewable always apply. Show built-in prompt shows what the helper sends.
- Studio AI 0.5.3: every suggested change arrives as a diff
In Agent mode, code the model writes in its answer is turned into proposed changes in the same turn, so you review diffs instead of copying code. Buttons under answers now say Draft, because nothing is written until you accept.
- Studio AI 0.5.2: Auto picks the model for you
Choose Auto and each message goes to a fast model for simple questions or a strong one for changes, reviews, fixes and images. If the chosen model is rate-limited or down, Auto tries the other. Ask mode now only lists fixes when you ask for a review or a fix.
- Studio AI 0.5.1: see the raw request and response of every turn
Tick Raw in the chat view and each answer gets a collapsed section with every model call it made: method, URL, HTTP status and the exact request and response bodies. Headers and API keys are never recorded.
- Studio AI 0.5.0: passwords and secrets scrubbed before the model sees them
Passwords, API keys, tokens and client ids in properties, YAML, Mule XML, URLs, console output and your messages are hidden from the model as «masked-N». Proposed edits still apply with the real values. .env files, private keys and keystores are never sent.
- Studio AI 0.4: proposal checks, connector lookup and Send to MuleSoftForge Studio AI everywhere
Proposed edits are checked before you apply them, the helper can look up connectors, and you can send a selection to the chat from DataWeave editors and the flow canvas. 0.4.6 tidies the chat composer onto one row.
- MuleSoftForge Studio AI launches
An AI helper inside Anypoint Studio for Mule XML, DataWeave, Problems and Console output. Bring your own key for OpenRouter, MuleSoft AI Gateway, OpenAI, Anthropic, Ollama or any OpenAI-compatible endpoint.
JSON Logger documentationSet-up, the Logger operation and the Logger Scope are now documented, including forwarding to Anypoint MQ, JMS and AMQP.
TypeSafe Connector 1.0.3: Locate, Why a Connector and the playgroundLocate returns a bounding box for image questions, and a new sample app playground runs every operation from the browser.
- TypeSafe Connector 1.0.2: per-option Choice and three-band policies
Apply Policy now fails closed and validates policies, Choice supports per-option thresholds, and Yes/No policies gain a REVIEW band.
- TypeSafe Connector for Mule 4 on Maven Central
Typed decisions with Jev from a Mule flow: Evaluate, Ask Yes/No, Choose and Score with native DataSense.
Agent Network Studio and the Salesforce UserInfo Claims Enrichment PolicyA browser studio for MuleSoft Agent Networks, and a Flex Gateway policy that brings per-user Salesforce attributes to MCP tool authorization.
Informatica MDM Connector 1.0.5: non-blocking re-authenticationA 401 now triggers asynchronous re-login and retry, so the HTTP requester thread is no longer blocked.
- Informatica MDM Connector 1.0.4: MetaData Read and Http Request
Passthrough authentication, automatic reconnect on 401, a MetaData Read operation and the renamed Http Request operation.
Lettuce Redis Connector documentationEvery operation group, Pub/Sub sources and Send Command are now documented.
Mule Chunking Connector 0.2.0Released to Maven Central under com.mulesoftforge.
PDFBox Module 1.0.2: Image to PDFConvert images into PDF documents inside a flow.
IDP Connector 1.0.6Maintenance release on Maven Central.