E
EZRA
← All posts

Introducing EZRA v6

EZRA Team

Why we built this

The promise of AI coding agents was speed. The reality, two years in, is something more ambivalent: they ship a lot, but their work is hard to govern. Decisions evaporate between sessions. Standards drift across files. Costs creep. And nobody — not the reviewer, not the author, not the next agent — has a single source of truth for "what did we already decide, and is the codebase still consistent with it?"

EZRA v6 is the answer to that question. It is a codebase governance layer for teams that ship with AI. The CLI runs locally with zero dependencies; the dashboard renders the same state in the cloud so a team can monitor health, decisions, agent costs, and compliance from one place.

What's new in v6

  • 41 slash commands — `/ezra:scan`, `/ezra:health`, `/ezra:decide`, `/ezra:research`,
  • `/ezra:plan`, `/ezra:portfolio`, and 35 others.
  • 44 hooks — wired into the standard Claude Code events (`PreToolUse`,
  • `PostToolUse`, `UserPromptSubmit`, `SessionStart`, `Stop`, `SubagentStop`,
  • `Notification`, `PreCompact`).
  • 90 workflow templates across 16 categories — phase gates, code review,
  • release, hotfix, canary deploy, dependency audits, on-call handoffs, RFC drafts.
  • Plus 10 dedicated AI Resilience templates that wrap multi-agent runs with
  • budget caps, output review, and cross-agent consistency checks.
  • Multi-agent orchestration — Claude, GPT, Codex, Cursor, Windsurf, Ollama,
  • Copilot, and Gemini, with per-agent daily and monthly budgets and a shared
  • cost-and-quality ledger.
  • 30+ compliance standards — ISO/IEC/IEEE 12207, 25010, 27001, 42001 (AI
  • Management), the EU AI Act, NIST AI RMF, and more, scopable as Global or Project
  • with email-confirmed change approval.
  • Cloud sync dashboard — opt-in, read-only by design, with row-level security
  • scoped to the signed-in user.

Three minutes to first signal

npm install -g ezra-claude-code
cd your-project
ezra init && ezra scan

That's enough to get a 5-pillar health score (standards, security, testing, docs, architecture), a list of decisions captured during the scan, and a starter `.ezra/governance.yaml` your future agents will read on every action.

How EZRA stays out of your way

Hooks are zero-dependency Node scripts. They read JSON on stdin and write JSON on stdout. Non-zero exit denies the tool call; otherwise the agent's request goes through. EZRA never writes to your project source tree — everything it persists lives under `.ezra/`, source-controllable, safe to delete and regenerate.

The dashboard is the same story: it is read-only. No widget on the dashboard ever writes back to your local `.ezra/` state. The CLI pushes a redacted snapshot when you run `ezra sync`; the dashboard renders. That separation makes the cloud piece a "nice to have" rather than a hard dependency.

Pricing

  • Core is free, MIT-licensed, and includes the CLI, all 41 commands, all 44
  • hooks, and the full 90-template workflow library.
  • Pro ($29/user/mo) adds multi-agent orchestration, the cloud-sync dashboard,
  • the research agent, the memory system, and the planning engine.
  • Team ($59/user/mo) adds multi-project portfolio, cross-project learning,
  • team management, session handoffs, and advanced analytics.
  • Enterprise is custom — SSO/SAML, self-hosted, audit logging, SLA.

See the [pricing page](/pricing) for the full comparison.

What comes next

We're already talking to the first wave of teams about a few things we deliberately held back from v6 to keep the launch tight: a JetBrains plugin, a self-hosted distribution, and a richer on-call integration. If any of those would change your calculus, let us know — we read every email at `hello@ezradev.com`.

Get started

npm install -g ezra-claude-code
ezra init

Or [sign in](/login) to provision the cloud-sync dashboard. The Core tier covers the entire CLI surface — there's no paywall on the things that make EZRA useful day one.

Welcome to v6.