Galileo continued to sharpen its focus on AI agent governance and evaluation this week, centering activity around the open-source launch of its Agent Control platform. The control plane is designed to let enterprises centrally define and enforce behavior policies across fleets of first- and third-party AI agents without redeployments.
Claim 30% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
Agent Control decouples guardrails from individual agents, addressing what Galileo describes as brittle, hard-coded safety controls that hinder production scale. Released under the Apache 2.0 license, the vendor-neutral platform integrates via hooks with existing agent frameworks and supports multiple guardrail providers, including Galileo’s Luna models, NVIDIA NeMo, and AWS Bedrock.
The company highlighted early ecosystem support from partners such as Amazon Web Services, Cisco AI Defense, CrewAI, Glean, ServiceNow, Rubrik, and Strands Agents. These integrations suggest a go-to-market path into large enterprises and digital-native customers facing rising volumes of AI agents and token traffic as deployments mature.
Galileo underscored that centralized policy management aligns with emerging best practices in “eval engineering” and lifecycle-based continuous improvement of AI systems. Commentary from an IDC research director framed unified control planes and evaluation workflows as key enablers for accelerating time-to-value while managing risk and compliance in AI projects.
Complementing the open-source release, Galileo is promoting a webinar to showcase Agent Control’s capabilities in real time. The session, led by the co-founder and CTO along with product and engineering leaders, will feature a live demo focused on runtime steering and blocking of agent behavior, as well as the use of an @control() decorator for policy management.
The company is positioning the tool to empower both developers and non-technical teams, allowing policy updates within minutes via a centralized interface. This cross-functional approach could help embed Galileo’s solutions deeper into operational workflows and broaden its buyer base beyond core engineering stakeholders.
In parallel, Galileo advanced developer tooling and education around AI evaluation and observability. It extended its MCP Server and Signals monitoring integration into IDEs like VS Code and Cursor, aiming to connect detection, root-cause analysis, and automated code fixes directly in developer environments.
The platform also added support for Claude Sonnet 4.6 and Gemini 3.1 Pro across its Playground, Prompt Store, and Metrics Hub, and emphasized compatibility with Microsoft’s Agent Framework via OpenTelemetry. These moves are intended to align with enterprise observability standards and simplify trace logging across diverse AI stacks.
To deepen evaluation capabilities, Galileo introduced three new RAG metrics: Chunk Relevance, Context Precision, and Precision@K, alongside an Enterprise Beta for Annotation Queues to streamline human-in-the-loop review. These features target more rigorous monitoring of retrieval quality and structured expert feedback for production systems.
On the education front, Galileo released a free “Eval Engineering” book and partnered on a Udemy course with educator Henry Habib to standardize best practices in AI reliability. Together, these initiatives reinforce Galileo’s strategy to become core infrastructure for AI agent governance and evaluation, marking a strategically significant and active week for the company.