Building the Trust Layer for the Agentic Era | Sonar Summit 2026

The Adaptive Developer in an AI-Driven World

As software development enters the agentic era, organizations face unprecedented change in how code is written and maintained. Ori Yitzhaki, Chief Product Officer at Sonar, emphasized that success in this transformation belongs not to the strongest or smartest developers, but to those who can adapt fastest. The reality is that different organizations and developers approach AI at different paces—some view AI as a copilot to assist human developers, while others have fully embraced autonomous agents capable of generating hundreds of lines of code in seconds. Sonar's mission is to support organizations across this spectrum, ensuring that the platform remains relevant regardless of their chosen development approach.

The Agent Hierarchy: From Planning to Problem Resolution

Effective agent-driven development requires a structured methodology that Sonar describes as a "muscle hierarchy" for agents. Before triggering any coding agent, developers must complete two mandatory stages: planning and context provision. Planning involves guiding agents with clear specifications and objectives, while context encompasses code understanding, guardrails, and architectural constraints. Once agents generate code, a two-cycle verification process becomes essential—first validating the agent's reasoning to ensure it understands its task, then verifying the actual code output. Finally, discovered issues must be autonomously resolved before deployment. This framework, represented by the guide-verify-solve traffic light model, forms the foundation of Sonar's roadmap for trustworthy AI development.

Context Augmentation: Solving Agent Amnesia

Agents suffer from a critical flaw: amnesia. Without proper context, agents guess, sometimes inventing fake APIs or referencing deprecated libraries that developers abandoned years ago. Sonar's solution is context augmentation—a new service that makes agents aware of existing code issues, security flows, and architectural constraints within the source codebase. Context extends far beyond simple prompts; it encompasses code, documentation, tickets, and structural information, while also accounting for dynamic elements like false positive markings within SonarQube. Currently exposed through empty files compatible with tools like Cursor, this context will soon be accessible via Sonar's MCP (Model Context Protocol) server for direct integration with IDEs and cloud development environments. Benchmark testing across multiple open-source projects in Java, C, and Python demonstrated a 66% reduction in issues created by agents when provided with proper context augmentation.

Preventing Architectural Drift with SonarQube Architecture

A significant challenge in agent-driven development is architectural drift—when AI-generated code deviates from intended system architecture, potentially causing maintenance nightmares, system instability, and security vulnerabilities. To address this, Sonar introduced SonarQube Architecture, a new product that automatically discovers and validates architectural integrity. The system uses an architectural discovery engine to build a graph representation of application components, allowing developers to define intended architecture and rules. During each scan, SonarQube Architecture compares existing code to the intended design and flags violations as quality issues, enabling teams to maintain architectural consistency alongside code quality and security compliance.

Supporting Diverse Development Strategies

Rather than prescribing a single path forward, Sonar acknowledges that organizations will adopt agent-driven development at their own pace and according to their own needs. The platform provides comprehensive tools whether teams use AI as a copilot for human-assisted development or deploy fully autonomous agents. The MCP server integration offers a straightforward interface for connecting agents directly to SonarQube, making it convenient for development teams to begin their agentic journey. By addressing fundamental challenges like context awareness, agent reasoning verification, and architectural compliance, Sonar aims to build trust in AI-assisted development practices while maintaining the flexibility organizations need during this transformational period.

Key Takeaways

• Agent success requires context: Agents without proper codebase context generate unreliable code; context augmentation addresses this through structured, dynamic information from SonarQube.
• Verification is mandatory: Both agent reasoning and code output must be verified before deployment to ensure quality and security.
• Architectural drift is costly: SonarQube Architecture automatically detects when AI-generated code violates intended design patterns, preventing maintenance and security issues.
• Flexibility supports adoption: Organizations can adopt agent-driven development at their own pace, whether using AI as a copilot or autonomous agent system.
• Context augmentation delivers measurable results: Benchmark testing showed 66% reduction in generated issues when agents received proper contextual information.