All notable changes to ArqZero are documented in this file.
The format follows Keep a Changelog, and this project adheres to Semantic Versioning.
- Multi-provider expansion: Anthropic, OpenRouter, OpenAI-compatible adapters
- Provider factory + registry layer (
src/api/factory.ts,src/api/registry.ts) - OpenRouter multi-key fallback chain — auto-rotates on
401/402/403/429/5xx /providerslash command for live provider switching
- README: license badge corrected (Apache 2.0), photo path fixed
- Stale launch-priorities HTML artifact (
docs/research/arqzero-launch-priorities.html)
2.0.0 — 2026-03-22
- Commercial layer — backend (Hono + Supabase + Drizzle), website (Next.js, GitHub Pages), CLI auth
- Email-based device auth (Resend) with brute-force protection + 5-attempt rate limit
- JWT access tokens (1h) + refresh tokens (90d) with 7-day offline grace
- Distributed feature gating across tools, commands, and capabilities
- Path-guard with sensitive directory blocklist (
.ssh,.gnupg,.aws,.env) - Worktree isolation flag (
--worktree) - Cross-session memory at
~/.arqzero/memory/ - 11 hook event types (PreToolUse, PostToolUse, SessionStart, SessionEnd, etc.)
- Plugin manifest + loader + manager
- Dispatch tool — up to 7 parallel subagents with auto model routing
- 18 built-in tools, 42 structured capabilities
- 25 slash commands incl.
/compress,/check,/setup,/think,/undo,/save-session - Verification gates (build, type, lint, test, security, diff) before completion claims
- Custom slash commands via
.arqzero/commands/*.md - Strategic context compaction at token limits
- Comprehensive test suite — 747 CLI tests, 24 backend tests
- Default provider: Fireworks AI with GLM-4.7 (Enso) and GLM-5 (PRIMUS)
- TUI rebrand to "The Grid" — amber/gold (#FFB800), shimmer spinner, command center
- Config dir moved to
~/.arqzero/ - npm package renamed to
arqzero, bin commandarqzero - License switched to Apache 2.0
- Hard startup error if
JWT_SECRETmissing or < 32 chars crypto.randomIntfor verification codes (replacedMath.random)- Rate limiting on all auth endpoints (login, verify, refresh)
- CORS conditional on
NODE_ENV(no localhost in production) - Old token cleanup on login + verify
- Team-memory value cap at 64 KB
- Old
/auth/verifyendpoint invalidates all unused tokens after 5 wrong attempts
Pre-2.0 was the original Claude Code clone phase. v2.0 was a rewrite with the labs-native architecture, commercial layer, and Fireworks default.