Confluent announced quarterly updates to its Confluent Intelligence and AI Tools portfolio. Highlights include GA of Real-Time Context Engine with new usage/cost metrics, early access support for IBM Granite time-series models (TTM, FlowState, PatchTST-FM) and Google TimesFM for anomaly detection and forecasting via Flink SQL functions AI_FORECAST and AI_DETECT_ANOMALIES, a GA fully managed MCP server with new write tools for connector restart/config updates, new open-source Agent Skills (Migrate, Java Kafka client, Flink UDF, Skill Creator) for coding assistants like Claude Code and Cursor, and an early-access Confluent Copilot for natural-language troubleshooting and operation of Confluent Cloud, including automated connector diagnostics and fixes.
Table of contents
Q3’26: What’s New in Confluent Intelligence and AI ToolsReal-Time Context Engine (GA): Give AI a Live View of the BusinessExpanded Model Support for Real-Time Anomaly Detection and Forecasting: IBM Granite Time-Series Models (EA) and Google TimesFM Model (EA)Updates for Fully Managed MCP ServerNew Agent SkillsNew Confluent Copilot: Confluent Expertise, Available Through Natural LanguageBuild Production-Ready AI with Confluent Intelligence and AI ToolsQuestions this post answers
How do I use IBM Granite time-series models or Google TimesFM for anomaly detection in Flink SQL?
Both are selectable directly through Confluent's existing AI_FORECAST and AI_DETECT_ANOMALIES Flink SQL functions using a single SQL parameter to pick the model, with no pipeline redesign required. IBM Granite models (TTM, FlowState, PatchTST-FM) and Google TimesFM are available in Early Access on AWS, hosted in Confluent Cloud with zero setup or credential management. daily.dev surfaces streaming and AI tooling updates for teams evaluating forecasting models in production.
What new capabilities does Confluent's fully managed MCP server support?
The fully managed MCP server, now generally available, adds write tools for restarting connectors and updating connector configuration, with Flink tools for creating and managing Flink SQL coming later in September. It gives AI assistants permissioned access to Confluent Cloud resources without teams needing to deploy or maintain their own MCP infrastructure. Developers wiring AI assistants into cloud infrastructure track MCP server capabilities on daily.dev.
What are Confluent Agent Skills and which ones were released?
Confluent Agent Skills are open source packages of Confluent expertise for AI coding assistants like Claude Code and Cursor, installed via npx skills add confluentinc/agent-skills. New skills include Migrate (MSK to Confluent Cloud migration planning), a Java Kafka producer/consumer client scaffolder, a Flink UDF builder, and a Confluent Skill Creator for building additional skills. daily.dev helps developers adopting AI coding assistants keep pace with new agent skill releases.