Baz
Also known as: Baz, Baz Technologies, Baz Technologies, Inc., baz.ai, baz.co, baz-scm, Baz Planner, Spec Reviewer, Awesome Reviewers
Engineering review platform whose agents govern AI-written code end to end: an AI code reviewer, a Spec Reviewer that runs your app in a sandbox and drives a browser to check the built experience against Figma and Jira, and Planner, whose plans a person approves before code is written.
Baz, based in Boston, is an agentic coding platform whose AI agents operate directly on a team's codebase to review, govern and secure the code that developers and AI copilots produce. It was founded by former Palo Alto Networks engineers who helped build that company's cloud application security business. Chief executive Guy Eisenkot earlier co founded Bridgecrew, which Palo Alto Networks acquired. In June 2026 Baz extended its seed round to $17 million, co led by Battery Ventures and Boldstart Ventures. Its premise is that in a world where AI generates code faster than any team can review it, reviewing after the fact is no longer a viable primary defense.
Baz began with an AI Code Review agent that runs on every pull request. It reads the codebase, detects breaking changes across endpoints and behavior, and enforces standards across product, design, architecture, security and reliability. In October 2026 five core review agents replaced eleven narrower reviewers, and a dedicated agent security reviewer began running on changes that touch agent code such as MCP servers, skills, hooks and CI workflows that hold model credentials.
Its Spec Reviewer agent pulls requirements from Jira and other ticketing tools and designs from Figma, then launches the application in a sandbox and drives a browser to check that the delivered experience matches intent. Baz Planner moves review upstream. It produces an implementation plan that goes through an approval workflow, with versioning and comments, before any code is written, and Merger returns a merge, reject or escalate verdict after CI passes, approving or merging only where a rule allows it. Reviewers learn from repeated team feedback and update their own prompts. Session Logs show how each run was triggered, where it ran, the stages it completed, its outcome and its cost.
Baz offers AI code governance that gives engineering and security leaders control over a codebase increasingly authored by systems they cannot fully audit. It acts as a super harness that coordinates coding agents rather than replacing them. Its reviewers run on OpenAI's GPT-5-Codex with Anthropic Sonnet 4.5 reflections, and Baz chooses the models. Private Mode keeps source code in the customer's own VPC, and Enterprise adds self hosted deployment in the customer's own cloud. Baz is built for teams shipping with AI copilots, with code review that learns from the team, spec validation and pre code governance integrated into GitHub and the IDE.
From our AI code reviewers video
AI Code Review Tools: Which Ones Run the Code They Review?
Our video on the 7 AI code reviewers in the index names what each one does best. For Baz, that is checking the built thing against what was asked for.
The player starts at Baz's part, 5:14 in. Recorded in October 2026; the profile here is live and updates as research is verified.
Vendor details
Canonical URL
https://baz.ai
Category
Coding agent
Subcategory
Agentic AI code review and governance
Funding status
Independent, based in Boston, founded by a team of former Palo Alto Networks engineers who helped build its cloud application security business, and led by chief executive Guy Eisenkot, who earlier co founded Bridgecrew before its acquisition by Palo Alto Networks. Baz Technologies raised an eight million dollar seed co led by Battery Ventures and Boldstart Ventures, then extended it by nine million dollars to seventeen million dollars total in June 2026, with participation from AFG Partners and Disruptive VC. Its AI Code Review tool ranks first on the precision weighted Code Review Bench, and it is available on the GitHub and AWS marketplaces.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Baz connects to GitHub, GitLab and Azure DevOps for source control. GitLab supports two way comment and merge sync through a personal access token, and Azure DevOps has project scoped onboarding. Ticketing integrations cover Jira, Linear, Azure DevOps, Monday, Shortcut, YouTrack and Fibery, with acceptance criteria seeding the spec and optional write scopes to post conclusions back. Figma connects through a personal access token for design, Notion and Google Docs supply engineering knowledge, Datadog and Sentry feed the SRE agent, and Slack carries chat. Baz works alongside Claude Code, OpenAI Codex, Cursor and Devin as a super harness that coordinates coding agents rather than replacing them. Surfaces include the pull request, a PR Inbox, an interactive CLI review loop, an MCP server for IDE agents, and a slash command that starts a Planner session from inside another coding agent. Baz is distributed through the GitHub and AWS marketplaces.
In practice
Your AI copilots generate diffs faster than anyone can review, and risky changes slip into production. Baz reviews every pull request with agents that understand your codebase, flag breaking changes, and enforce your standards.
Code compiles and passes tests but silently drifts from the Figma design and Jira acceptance criteria. Baz's Spec Review agent pulls the requirements and uses browser automation to verify the delivered experience matches intent before merge.
You want to stop reviewing AI code after the fact and catch problems earlier. Baz Planner reviews the plan against your architecture and constraints before code is written, and a reviewer approves the plan or requests changes before implementation starts.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
13.0 / 14 capabilities · 93%
| Integrations & Tool Calling | Full |
|---|---|
|
Integrations cover GitHub, GitLab and Azure DevOps for source control; Jira, Linear, Monday, Shortcut, Azure DevOps Boards, YouTrack and Fibery for tickets; Figma for design; Notion and Google Docs for knowledge; Datadog and Sentry for the SRE agent; Slack for chat; and preview environments and a sandbox for running the app. Sourcebaz.ai/docs/basics/integrationsread 2026-09-29 |
|
| Workflow Orchestration | Full |
|
Multiple agents divide the work on a change: SDLC agents (Fixer, Merger, SRE, Skill Maintainer) run independently in sandboxed environments alongside a set of individually scoped review agents, Merger decides after CI passes whether a pull request is safe to merge, and Planner produces an implementation plan that goes through approval before code is written. Sourcebaz.ai/docs/agents/baz-agents and baz.ai/changelog 2026-07-01read 2026-09-29 |
|
| Knowledge Grounding & RAG | Full |
|
Reviewers ground on the organization's repositories through API endpoints for file search, text grep and cross repository architecture search. They also take product intent from ticketing, Figma, Notion and Google Docs integrations, and the Spec Reviewer checks the code against it. Sourcebaz.ai/docs/account/api-keys, docs/basics/integrations and docs/agents/baz-agents; baz.ai/docs/basics/integrationsread 2026-10-25 |
|
| Human Oversight & Guardrails | Full |
|
Plans generated by Planner go through a dedicated approval workflow before implementation starts, and teammates can review plans together and track revisions, so a person approves before the agent's plan is built. Merger returns a merge, reject or escalate verdict after CI passes and labels the pull request; it approves only when a rule turns on auto approve, and merges only when configured to act autonomously, otherwise the verdict is left for a human reviewer to act on. Sourcebaz.ai/docs/agents/baz-agents and baz.ai/changelog 2026-07-01 and 2026-09-24read 2026-10-28 |
|
| Security, Identity & Governance | Full |
|
Baz states that it is SOC 2 certified, Enterprise adds SSO, centralized controls and VPC deployment, and API keys are limited to a whitelist of endpoints under the creator's permissions. Sourcebaz.ai/docs/account/security-privacy-and-compliance, baz.ai/pricing and docs/account/api-keysread 2026-09-29 |
|
| Observability & Auditability | Full |
|
Session Logs let a team inspect how each run was triggered, where it executed, its status, the major stages it completed, its outcome and its cost, and the API records MCP session events. Sourcebaz.ai/changelog 2026-06-01 and docs/account/api-keysread 2026-09-29 |
|
| Memory & State Persistence | Partial |
|
Reviewer Memory stores a memory in a reviewer's own memory bank when similar feedback is given more than once, and updates that reviewer's prompt. Each memory can be switched on or off and shows its sources. An organization setting can hold new memories for admin approval before they go live, and admins control who may contribute. No lifetime, expiry or delete path is stated, so a disabled memory stays in the bank. Sourcebaz.ai/docs/agents/working-with-agentsread 2026-10-25 |
|
| Deployment & Data Residency | Full |
|
Private Mode, on Pro and Enterprise, keeps source code stored only in the customer's own VPC through a file system service pod on AWS EKS, with analysis on Baz infrastructure reading it transiently. Enterprise adds self hosted deployment in the customer's own cloud, and private application review can run the sandbox inside the customer's Kubernetes as a limited preview. Sourcebaz.ai/docs/account/private-mode and baz.ai/pricingread 2026-10-25 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Baz ships named built in agents with their own jobs, among them Fixer, Spec Reviewer and an AI SRE agent. Five core review agents (Code Correctness, Code Quality, Basic Security Patterns, Advanced Security and AI Coding Guidelines) replaced eleven narrower reviewers in October 2026, and teams can build customized agents on top. Sourcebaz.ai/docs/agents/baz-agents and baz.ai/changelogread 2026-10-25 |
|
| Triggers & Channel Coverage | Full |
|
Reviews run automatically when pull requests open across GitHub, GitLab and Azure DevOps, Merger runs after CI passes, and Skills Maintainer can scan on a schedule every three days, weekly or every two weeks. Comment commands, the CLI and the MCP server add runs on demand. Sourcebaz.ai/changelog 2026-08-10 and docs/agents/baz-agentsread 2026-09-29 |
|
| Model Flexibility & Routing | Partial |
|
Baz's agentic reviewers run on OpenAI's GPT-5-Codex with Anthropic Sonnet 4.5 reflections, and Baz chooses the models, with no customer model selection, bring your own key or local model option offered. Sourcebaz.ai/changelog 2025-10-16read 2026-09-29 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
API keys call the Baz REST API (v2, under /api/v2/ with an x-api-key header) and the Baz MCP server for scripts, CI and unattended agents, with twelve endpoints covering reviewer configuration, review discussions, repository search, plan creation and versioning, plan comments and linking pull requests to plans. Sourcebaz.ai/docs/account/api-keysread 2026-09-29 |
|
| Testing, Debugging & Optimization | Full |
|
Prompt Playground lets a team test a review agent's behavior on real change requests before rollout and tune its tone, scope and strictness without pushing untested agents to production, and Evaluations report reviewer performance with usage metrics and comment tracking. Sourcebaz.ai/changelog 2025-05-22 and 2025-07-29read 2026-09-29 |
|
| Browser & Computer Use | Full |
|
Spec Reviewer clones the repository, restores dependencies, runs the setup and start commands, waits for the application to become healthy and opens it in a browser to check the built experience against the specification. Sourcebaz.ai/changelog 2026-07-29read 2026-09-29 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Starting November 3, 2026, Baz Pro moves from seats plus Engineering Work Credits to $30 per developer each month with 50 reviews included, then $1 for each additional review. Agent work such as Fixer, Advanced Security and AI SRE is now billed by usage in dollars instead of credits. Accounts switch at their next billing cycle, and signed annual agreements keep their contract terms.
Bears on: Pricing / packaging
View sourcegion is a new open source git client from Baz, built for coding agents and CI jobs and now in beta. It fetches only the files, trees and commits a task asks for, so an agent can read, search, commit, push and rebase on a huge repository without a full clone. One command installs a skill that has a coding agent use gion in place of git clone, and a GitHub Action gives CI jobs a sparse checkout.
Bears on: MCP / tool calling / API
View sourceA dedicated agent security reviewer in Baz now runs on code changes that touch agent code, such as agent configuration, model SDK usage, direct provider calls, plugins, skills, hooks, MCP servers and CI workflows. It works from a curated reference of agent specific threats, including untrusted context flowing into prompts, bypassable hooks and guardrails, secrets exposed through skills, an unpinned agent supply chain and unauthenticated agent and tool surfaces. Each finding must rest on evidence and appears as a line comment tagged with its OWASP Top 10 for LLM Applications category.
Bears on: Security / enterprise
View sourcePricing
Pro from $30 per active developer per month, plus Engineering Work Credits at $0.01 each for advanced agents. Enterprise is priced on contact.
Per active developer per month for access and standard review, plus Engineering Work Credits at $0.01 each for advanced agent sessions.
What is public
Pro is $30 per active developer per month after a 14 day free trial, with standard AI code review included. Engineering Work Credits cost $0.01 each for advanced agents. Enterprise is priced on contact and adds self hosted deployment in the customer's own cloud, SAML and OIDC SSO, centralized admin controls and custom terms.
Billing mechanics
Seats cover access and standard AI code review. Fixer, Advanced Security, Spec Reviewer and AI SRE sessions draw Engineering Work Credits at $0.01 per credit, with published typical ranges per session and a suggested budget of $20 to $50 per active developer per month.
Cost watchouts
Engineering Work Credits for advanced agents are billed on top of the per developer seat price.
Variable cost rationale
Advanced agent sessions are metered. Fixer, Advanced Security, Spec Reviewer and AI SRE draw Engineering Work Credits at $0.01 each, with typical sessions costing from $1 to $4.30, so the bill grows with how often those agents run. The seat price per active developer is fixed and covers access and standard AI code review, which draws no credits. The suggested planning range for credits is $20 to $50 per active developer per month.
Additional watchouts
The seat price covers standard review only. Advanced agent sessions are billed in credits on top, so heavy use of Fixer or Spec Reviewer raises the monthly cost.
Overage / add-ons
Advanced agent sessions consume Engineering Work Credits at $0.01 per credit, with typical sessions from $1 to $4.30 for Fixer, $1.75 to $2.20 for Advanced Security, $1 to $2.50 for Spec Reviewer and $1.50 to $3 for AI SRE.
Sales call required
Mixed (some tiers require a call)
Free / trial
14 day free trial of Pro with full platform access, and no permanent free plan
Lowest paid plan
Pro at $30 per active developer per month, with Engineering Work Credits billed on top.
Key ambiguities
Enterprise pricing is not published, and credit spend depends on how often advanced agents run.
Cancellation / refund
The trial runs 14 days, then Pro bills $30 per active developer per month. Enterprise has custom terms.
Support SLA / resale
Pro includes community and email support, and Enterprise adds dedicated support and onboarding. Enterprise is also sold through AWS Marketplace.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- 10Web — Agentic website platform whose AI agents build, host and manage…
- AgentUI — Managed AI app builder for operations teams that generates and hosts…
- Aider — Open source, model agnostic terminal coding agent that edits your…
- Amp — Amp is a coding agent from Amp Frontier Corporation that runs each…
- Anthropic Claude Code — Anthropic's agentic coding system across terminal, desktop, IDE, web…
Alternatives to Baz
The closest documented capability profiles to Baz among coding agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Zencoder13.5 / 14Fuller documented coverage on Model Flexibility & Routing
- Factory13.0 / 14Fuller documented coverage on Model Flexibility & Routing
- Lovable12.0 / 14A lighter documented profile than Baz
- Warp13.0 / 14Fuller documented coverage on Model Flexibility & Routing
- Anthropic Claude Code12.5 / 14Fuller documented coverage on Memory & State Persistence
- Capy12.5 / 14Fuller documented coverage on Model Flexibility & Routing
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded