Baz

Domain-expert agents on every code changes

These agents are designed to provide specialized knowledge and insights on code changes, ensuring that every modification aligns with best practices and project requirements. Users interact with these agents to receive tailored feedback and guidance during the development process.

Spec Reviewer

Validates that product and design requirements were actually implemented by comparing pull request behavior against tickets, specs, and observable outputs. This feature ensures that the development aligns with the intended design and functionality, providing confidence in the code changes.

Logical & Breaking Change Detection

Identifies deep regressions, unsafe API shifts, and deviations from architecture patterns. This capability helps developers catch potential issues early in the development cycle, reducing the risk of introducing bugs into production.

Fixer Agent

Executes deterministic, verifiable fixes in a controlled environment. Fixes are audited, reproducible, and CI-safe, allowing developers to trust that the changes made will not introduce new issues.

Track Acceptance & Impact

Unified telemetry measures acceptance rates, memory growth, and engineering impact over time. This feature provides insights into how code changes are received and their long-term effects on the codebase.

Multi-Source Awareness

Ingests tickets, design systems, observability data, CI signals, and preview environments to provide a comprehensive view of the development context. This feature allows developers to make informed decisions based on a wide array of relevant information.

Language-Aware Analysis

AST-based understanding across languages to reason about structure, not just text. This capability enhances the accuracy of code analysis by considering the syntactical and structural elements of the code.

Spec-Driven Validation

Ensures implementation details match documented intent, providing a layer of verification that the code adheres to the specified requirements. This feature helps maintain alignment between development and design specifications.

Production Signals

Correlates failures, regressions, and risk indicators to specific code changes. This feature allows teams to quickly identify the impact of changes on production systems, facilitating faster troubleshooting and resolution.

Adaptive Learning

Extracts rules from reviewer feedback and encodes them into structured, versioned prompts. This feature enables the system to learn from past interactions, improving its recommendations and insights over time.

Persistent per Repo & Org

Good memory should feel invisible. When memory is correctly scoped, relevant, and selectively applied, it simply removes noise, allowing developers to focus on what matters most.

Expert-Governed

Memory is auditable, editable, and tied to trusted maintainers. This ensures that the knowledge encoded in the system is reliable and can be updated as necessary to reflect current best practices.

Compounding Impact

As trust grows, teams act on an increasing percentage of Baz feedback, accelerating review cycles. This feature highlights the growing efficiency and effectiveness of the system as it learns and adapts to user needs.

Isolated sandboxes that execute code, process data, and run tools

These sandboxes provide a secure environment for executing code and processing data without affecting the main codebase. Users can experiment and test changes safely, ensuring that their work does not introduce errors.

Git

Clones Git repos into sandboxes to securely access specific branches, authenticate and perform git operations. This feature allows developers to work with version control in a safe and isolated manner.

Browser

Run and debug web applications in a full browser environment. This capability enables developers to test their applications in a realistic setting, ensuring that they function as intended.

Log & trace streaming

Allows agents to access and process logs and traces from any runtime and correlate them to build artifacts. This feature enhances the ability to debug and analyze code by providing visibility into runtime behavior.

Headless coding

Read and write code snippets in multiple languages for both stateless execution and stateful interpretation with context. This feature supports a flexible coding environment, accommodating various programming needs.