Spec Reviewer Agent
Validates code against product requirements, designs, and expected behavior - catching gaps and deviations before they ship.
Baz ensures your team builds software the way you intended.
AI coding is foundational to how software gets built, and teams need a new verification layer - Baz is defining it.
Baz is more contextual, scalable, and dependable - refreshing for teams shipping AI-native software.
Agents
Baz’s purpose-built coding agents run on top of the platform and operate like experienced engineers - in the context of behavior, requirements, APIs, architecture, production systems, and security.
Validates code against product requirements, designs, and expected behavior - catching gaps and deviations before they ship.
Reasons across auth and network boundaries, infrastructure, pipelines, and application code to uncover vulnerabilities.
Correlates repository changes with production telemetry to identify reliability, performance, and observability risks - then proposes fixes.
Automatically applies and validates safe code changes in an isolated runtime, turning review feedback into tested commits.
Planner
Baz Planner is a gateway between developers and the codebase. It routes every idea through dynamic loops that detect, root-cause, patch, and validate issues - rewriting coding plans to eliminate entire classes of bugs and vulnerabilities before they reach production.
retry_queue table and indicesSchema · safe to buildPOST /webhooks/retry endpointBroken Access Control · blockedEvery ad-hoc change is evaluated against current and prospective architecture and flagged before any code is saved.
Planner scores every model suggestion, blocks unsafe paths, and enforces the boundaries you define before progression toward production.
Teams report up to a 65% reduction in downstream rework, measured by revert and hotfix PRs following a merge.
What's new
At AI Engineer World's Fair, Baz announced Baz Planner - a gateway that eliminates entire classes of bugs and vulnerabilities in code planning - alongside $9M in new funding, bringing total raised to $17M.
How Baz uses Datadog LLM Observability to run autonomous agents safely in production - with end-to-end trace coverage, human-in-the-loop validation, and ~80% faster root cause analysis.
Six classes of vulnerability slipped past traditional tooling for years. A study of 28 high-confidence findings from advanced security review shows why cyber-capable models now catch them at review time - before they ship.
Session Logs give every Baz run a structured execution record. They show how a run was triggered, where it executed, its current state, major completed stages, outcome, and cost. Instead of exposing raw internal logs, Baz presents a timeline of product events so teams can inspect reviewer runs, fixer runs, scheduled scans, benchmark runs, and future Codex or Claude Code workflows from one place.
We're hiring
Help us build the agentic coding platform for AI-native engineering.
