What’s New in Arthur – August 2025 Edition
Agentic monitoring, stronger guardrails, and faster integrations
August was all about unlocking agentic workflows and tightening reliability across the Arthur stack. This month’s updates introduce an early look at agent monitoring in the Arthur Engine, and security and performance enhancements.
This month’s updates are designed to answer the questions our customers ask most often: What can I actually do with Agentic AI? How does it help me? Is this faster, safer, and easier than before?
Let’s dive into the story behind the latest features.
Why Agentic AI Matters
Imagine you’ve deployed an AI agent that books travel for your team. It can search flights, compare hotels, and even process payments. That’s powerful, but also risky. What happens if the agent goes off-script? Books a $12,000 ticket instead of $1,200? Pulls private data into the wrong system?
This is where Arthur’s Agentic AI support steps in.
With Agentic monitoring in the Arthur Engine, you can now track, evaluate, and compare agent behavior the same way you do with models. Instead of wondering if your agent made the “right” choice, you can monitor:
- Decision paths: Which actions did the agent take, and why?
- Reliability: Does it consistently follow your guardrails?
- Performance: Is it helping users reach goals faster, or slowing them down?
This isn’t just evaluation, it’s trust at scale. Teams can now experiment with agent workflows knowing they have a safety net to catch errors with better tool selection and response relevance before users feel the pain.
New Agentic AI Features to Explore
- The Arthur Platform now supports the ability to visualize agent traces and understand performance for key metrics at scale that unlocks your ability to understand agent performance and iterate with confidence.
New ODBC Database Connector: Less Plumbing, More Building
For many of our customers, the first bottleneck isn’t the model. It’s the data pipeline. Every team has critical business data sitting in systems like MSSQL, PostgreSQL, Oracle, or MySQL. Until now, bringing that into Arthur required heavy lifting.
The new ODBC connector changes that.
- One simple configuration: Standardized fields make setup painless.
- Faster integrations: Get data flowing in minutes, not days.
- More secure: Fewer custom scripts mean fewer vulnerabilities.
The impact? A data science team at a financial services company recently used the ODBC connector to pipe transaction data from their Oracle database directly into Arthur. Instead of waiting two weeks for custom ETL, they were running hallucination checks and guardrail evaluations within hours. That’s more time spent improving models, less time spent wrestling with pipelines.
New Data Connectors To Save You Time
- New ODBC database connector for MSSQL, PostgreSQL, Oracle, and MySQL with standardized fields and easy config.
Other Performance Enhancements & Usability Improvements
- CloudFormation launch button: Spin up Arthur with your client ID pre-populated. What used to take an hour now takes a few clicks.
- More accurate hallucination checks: Structured LLM outputs give you sharper insights into where and why responses go off track.
Whether you’re building with vision, text, or agent-based AI, Arthur helps you:
- Stay in control with monitoring that catches issues before users do.
- Experiment faster with connectors and automation that remove setup friction.
- Build trust with guardrails that flex to your use case.
The shift to Agentic AI is here, and with these updates, Arthur makes sure you can harness that power responsibly.
Stay tuned for the next wave of updates is just around the corner.
SHARE
