June 2025 Summaries

Sentry's guide for legal teams evaluating their Seer AI Agent outlines the stringent criteria the company's legal department uses to assess AI tools, emphasizing data protection and compliance with customer commitments. The criteria include limited data use rights, in-product data controls, alignment with customer commitments, audited security controls, and transparency. Sentry ensures that data is only used to provide services, with strict prohibitions on data sharing and model training without consent, and offers administrative controls for managing AI features. The Seer AI Agent is compliant with SOC 2 Type 2 and ISO 27001, undergoes annual security tests, and aligns with data retention and locality commitments. Despite not offering IP infringement indemnity for AI-generated code, Sentry argues that Seer's modifications are unlikely to pose significant infringement risks due to their nature as functional code fixes. The guide encourages legal departments to independently assess AI tools while presenting Sentry's checklist as a robust framework for evaluation.

Sasha Blumenfeld introduces a significant upgrade to Sentry Agent Monitoring, designed to enhance the debugging of AI systems by providing comprehensive visibility into complex AI operations. The update addresses the unique challenges of debugging AI code, offering a unified experience that combines tracing, tool visibility, and model performance insights. By delivering full execution breakdowns and model performance details, Sentry equips developers with the tools to identify and resolve issues efficiently. The platform's capabilities include error tagging and grouping, tool analytics, and comprehensive tracing that connects AI-specific data to the entire application stack. The enhanced Sentry Agent Monitoring aims to simplify the debugging process by capturing the entire flow of AI and non-AI components, thereby allowing developers to pinpoint failures and their impacts accurately. The tool is compatible with Vercel’s AI SDK and OpenAI Agents for Python, with more integrations forthcoming, and is available to all Sentry users, providing a seamless way to manage complex AI systems and ensure robust performance.

Sentry's latest updates to Insights enhance the ability to monitor and debug system issues by allowing users to set up alerts directly from charts, a feature that helps identify and address potential problems without constantly monitoring dashboards. The updates include built-in anomaly detection for span metrics, available on Business and Enterprise plans, which flags unusual behavior based on historical trends without requiring manual threshold settings. Insights provide out-of-the-box views that highlight system components likely to cause customer issues, displaying performance and error trends from auto-instrumented spans such as database queries and API calls. Users can investigate issues more deeply by exploring charts, create alerts to receive notifications of deviations, and manage key metrics by copying charts to dashboards. These improvements aim to simplify the process of spotting and resolving issues, ensuring a proactive approach to maintaining application performance and stability.

Sentry has introduced an open beta feature that integrates AI-driven test generation and pull request (PR) reviews directly into GitHub workflows, aiming to enhance code quality and streamline the development process. By simply commenting "@sentry generate-test" or "@sentry review" on a PR, developers can invoke Sentry's AI to automatically generate unit tests that cover code changes or provide structured reviews with summaries, key technical changes, risk considerations, and suggestions for improvement. This system is designed to reduce the burden of manual testing and PR reviews while supporting developers in maintaining robust codebases, minimizing regressions, and ensuring comprehensive test coverage. The tool leverages context from the repository, related files, and internet searches when necessary, offering a seamless and integrated experience without the need for additional plugins or steps in the IDE. Sentry encourages users to try these features for free, providing feedback via email or community channels, and emphasizes their role in complementing rather than replacing human feedback in the development workflow.

Seer is Sentry's newly launched AI debugger designed to enhance code debugging by providing precise root cause analysis and proposing merge-ready fixes specific to applications. Unlike traditional code generation tools that often struggle with debugging, Seer leverages comprehensive context from Sentry, such as stack traces, commit history, and profiling data, to efficiently diagnose and resolve issues. Since its beta phase, Seer has analyzed over 38,000 issues with a 94.5% accuracy rate in identifying root causes, significantly saving developers' time. The tool can automatically suggest and implement fixes, reducing the need for manual intervention while maintaining developer oversight. Seer operates across distributed systems, effectively troubleshooting complex issues by analyzing telemetry data and utilizing advanced models from AI leaders like OpenAI and Google. Privacy is prioritized as Seer doesn't use data to train AI models without explicit consent, and it is available to Sentry users on a paid plan with a free trial option.

The Sentry Flutter SDK version 9.0 introduces several features to improve visibility and debugging of errors in Flutter applications. Session Replay for mobile allows developers to see a video-like reconstruction of user interactions before a crash, while Logs (now in open beta) correlates logs with crashes and performance issues. Feature flag support enables understanding which flags were enabled when bugs happened, and Native JS error support provides better visibility into Flutter Web apps using JavaScript interop. Release Health support gives insights into release adoption, stability, and crash-free sessions for Flutter Web apps. Improved linking between traces and errors also correlates issues with specific spans in the application code. These features aim to provide more context and help developers identify and fix issues more efficiently.

The Sentry Unity SDK provides production-ready Unity crash reporting and error monitoring for real gameplay across various platforms, including desktop, mobile, web, and console. It offers human-readable stack traces, player actions and engine events leading up to an error, optional automatic screenshots, and performance insights into your game via distributed tracing. The SDK supports Linux, macOS, Windows, Android, and iOS, with additional support for Nintendo Switch, Xbox, and PlayStation consoles. Sentry also provides release health tracking, capturing session data, allowing you to monitor crash-free users and sessions, track adoption trends, and identify regressions introduced by new builds. By integrating the Sentry Unity SDK into your Unity project, you can debug issues without digging through endless logs, capture valuable context when manually logging errors, and group repeated exceptions under a single issue, making it easier to diagnose and resolve problems faster.

To configure lightweight browser tracing for debugging at scale, you can fine-tune your sample rate and span collection to make your traces more focused and readable. The `tracesSampleRate` setting allows you to control what percentage of frontend traces are sent to Sentry, while `ignoreResourceSpans` enables you to turn off certain resource spans that contribute to a large trace volume. By adjusting these settings, you can reduce your span volume without lowering your sample rate, which is particularly useful for high-traffic websites or when answering specific questions about user behavior. Additionally, you can customize your auto-instrumentation by ignoring performance API spans or disabling long animation frames, depending on the questions you're trying to answer and the structure of your trace. By asking "what questions are you trying to answer?" and considering both sample rate and configuration settings, you can get the most out of browser tracing for debugging at scale.

The Sentry MCP server is an open protocol designed to standardize context bringing into LLM interactions, and it has been added by Sentry as a hosted/remote solution. The MCP server supports various features such as remote hosted mode, OAuth support for logging in via existing Sentry organizations, Streamable HTTP with fallback to SSE if the client doesn't support it, and 16 different tool calls that can be used to bring wide context from Sentry into LLMs. These tool calls include project information, issues, finding or creating projects, and even calling the AI agent Seer to generate root causes and fix issues. The MCP server is designed to be low-friction and easy to consume as a service, with support for OAuth and Streamable HTTP. It can be added to most providers that support OAuth by adding specific content to the MCP configuration file. The Sentry MCP server is intended to bring context into LLMs, allowing them to use tool calls such as `list_projects` to access project information and chain together different tool calls to kick off Seer's analysis run. While MCP is designed for bringing context into LLMs, it is not built for debugging, and tools like Seer are purpose-built for understanding why applications break and how to fix them. The Sentry MCP server has made significant progress in a short period of time and is expected to continue growing in popularity and functionality.