Release History
Changelog
What's new in A.L.I.C.E. — every version, every change.
v1.5.13
2026-05-07- changeMission Control cloud bridge retired — the `--cloud` family of flags, `lib/mission-control.mjs`, `bin/alice-cloud.cjs`, and `templates/mission-control-bridge/` are gone. Alice Hub is the web product going forward; license validation against `getalice.av3.ai` is unaffected
- fixStarter installer no longer ships a dangling Athena → `nadia` delegation reference; a new defensive registry filter drops any unknown `allowAgents` entries with a logged warning so future template/tier drift can't silently break delegation
- fixManifest writer is self-healing — legacy `missionControl` blocks strip themselves on first write, and `packageVersion` refreshes on every install instead of pinning to whatever wrote the manifest first
- fixTier prompt counts corrected — Starter shows "14 agents", Pro shows "22 more agents, 36 total"
- featAthena coding team gained Symphony fan-out support in templates — `subagents.cloneSelf.max = 4` for Sasha/Dylan/Morgan/Priya/Felix/Quinn/Devon; `project_workspace` tool granted to all 14 starter + 22 pro agents; AGENTS.md docs for orchestration loop and specialist project participation
- choreTest suite covers the new defensive filter (Starter drops `nadia`, Pro keeps her) so starter↔pro template drift is caught at CI
v1.5.12
2026-04-29- featHermes Agent and OpenClaw/NemoClaw installs now use the same 33-agent roster, so the published product matches the documented team across runtimes
- fixInstaller state is more reliable across upgrades — runtime choice is persisted, `--force` now refreshes the managed install predictably, and Hermes uninstall cleans up tracked memory directories by installed roster
- fixPublished npm package is now privacy-safe by default — the public tarball ships only generic shared workspace templates, removing local user-history and machine-specific content from the package
- chorePacked artifact coverage expanded to verify Starter and Pro installs from `npm pack` and to block accidental reintroduction of local workspace files into the tarball
v1.5.11
2026-04-22- featRuntime auto-install wizard — guided setup when no runtime is detected
- featInstall Hermes Agent (~2 min), OpenClaw (~3 min), or NemoClaw (~20-30 min) from the installer
- featNemoClaw first-class support — proper detection, Docker pre-flight, official install command
- fix`detectRuntime()` now returns `null` when nothing found instead of silently defaulting to OpenClaw
- fix`--runtime X` offers to install X if it's missing
v1.5.10
2026-04-22- featHermes Agent install path — A.L.I.C.E. runs natively on Hermes as a skill tree
- featRuntime auto-detection (Hermes first, then NemoClaw, then OpenClaw)
- feat`--runtime hermes|openclaw|nemoclaw` flag to force a specific runtime
- featHermes skill templates for orchestrator + 33 specialists
- fixRemoved model selection wizard — uses runtime's configured model
- fixFail-fast with clear instructions if no model is configured
- docsREADME rewritten for Hermes-first, runtime-agnostic positioning
v1.5.9
2026-04-11- featAgent roster expanded from 30 to 33 specialists — MaxxiPro, AccuScope, and SmokeTestAgent now included in Pro tier
- featMaxxiPro: roof replacement industry specialist — sales systems, dealer networks, and product operations
- featAccuScope: CRM business intelligence agent — job profitability, A/R aging, and lead source analytics
- featSmokeTestAgent: QA specialist for automated testing workflows
- featAll 33 agents now ship with complete 10-file workspace templates (SOUL, IDENTITY, AGENTS, USER, PLAYBOOK, FEEDBACK, LEARNINGS, TOOLS, MEMORY, HEARTBEAT)
- choreWorkspace templates fully synced with live deployed agents
- docsROSTER.md updated with full agent bios and directory
v1.5.8
2026-03-28- fixReplace `got` dependency with native `fetch` in alice-cloud CLI — no more "Cannot find package 'got'" crash during gateway registration
- fixalice-cloud CLI v1.1.0 — zero external HTTP dependencies
v1.5.7
2026-03-28- fixBanner now says "31 AI agents, one team" (was "10 starter agents")
- fixSummary shows `Cloud: enabled (setup pending)` when token was skipped, not misleading `Cloud: enabled`
- fixFinal output shows prominent "☁️ Cloud setup pending" with exact command when cloud auth is incomplete
- fixCloud dashboard URL shown in final output when cloud is enabled
- fixalice.av3.ai/login?cli=1 now correctly shows token after sign-in (React state fix deployed)
v1.5.6
2026-03-28- choreRebranded all user-facing "Mission Control" references to "A.L.I.C.E. | Control" across installer, CLI help, doctor, and README
v1.5.5
2026-03-28- fixPro license no longer invalidated during upgrade when revalidation endpoint is unreachable — previously validated licenses now honored with grace period
- fixCloud onboarding uses installer's own readline instead of spawning a competing stdin process — no more input conflicts
- fixUpgrade flow correctly shows `Cloud: enabled` in summary when cloud was previously configured
v1.5.4
2026-03-28- featCloud onboarding rewritten — replaces old ingest token prompts with browser-based Supabase OAuth
- featInstaller auto-detects gateway URL/token and registers with alice.av3.ai automatically
- featHeartbeat daemon setup offered during install — keeps cloud dashboard connected
- featUpgrade mode shows existing cloud status and offers re-auth option
- featSkip cloud setup prints clear `alice-cloud login && alice-cloud register` instructions
- fixRemoved all references to `mc_ingest_` and `mc_worker_` tokens from user-facing prompts
- fix`alice-cloud.cjs` exports functions for programmatic use (no CLI execution on import)
v1.5.3
2026-03-28- feat`alice-cloud` CLI bundled — `npx @robbiesrobotics/alice-agents` now installs both the agent framework and the cloud onboarding CLI
- feat`alice-cloud login` — authenticate with Supabase (interactive or `ALICE_SUPABASE_TOKEN` env var)
- feat`alice-cloud register` — auto-detect local gateway URL/token, register with A.L.I.C.E. Cloud
- feat`alice-cloud watch` — heartbeat daemon (30s interval, auto-restart gateway if down)
- feat`alice-cloud status` / `alice-cloud unregister` — check or remove cloud registration
- featSupabase Realtime relay — gateway events stream live to the dashboard (session updates, logs, spawns, cron triggers)
- featNew Vercel API endpoints: `/api/cloud/register`, `/api/cloud/status`, `/api/cloud/heartbeat`
- featWelcome screen shows both Local and Gateway connection modes
- fixTheme and language switching now works without page reload
- fixSupabase service role key corrected across all API routes
v1.5.2
2026-03-26- featAgent roster expanded from 28 to 31 specialists — Aria (Autonomous Research Agent) and Nate (n8n Workflow Automation Specialist) now included in Pro tier
- featAria: long-running autonomous research agent per project; spins up sub-agents, sets cron jobs, and improves over time
- featNate: n8n workflow automation specialist; designs, deploys, and maintains n8n workflows via REST API for the full A.L.I.C.E. stack
- choreLanding page agent grid updated to reflect 31-agent roster
v1.5.1
2026-03-22- featPro purchases now auto-provision Mission Control ingest + worker tokens at checkout
- featToken-based multi-tenant auth — each install's events isolated by team_id in Supabase
- featBridge plugin events now flow live to alice.av3.ai (end-to-end verified: ingest accepted, node registered)
- featAdmin portal org PATCH endpoint — rename, add notes, update status on any organization
- featMission Control Supabase backend fully wired in hosted mode — sessions, agents, nodes all persisted
- fixMission Control env vars correctly set in Vercel (Supabase URL, anon key, service role key, admin token)
- fixContact form `/api/contact` route now live — was returning 404 previously
- fixAll 28 agent workspace templates now have bespoke SOUL.md — no generic placeholders
- docs10-page documentation suite live at getalice.av3.ai/docs
v1.5.0
2026-03-21- fixMission Control agent creation now accepts any model string — removed hardcoded enum that blocked non-standard models
- fixLLM service no longer forces a fallback model — passes through the configured model to the gateway cleanly
- fixAgent service defaults to null instead of hardcoded Gemini for model and provider when DB value is unset
- fixEmbedding service model names now configurable via `EMBEDDING_MODEL` env var instead of hardcoded
- fixTelemetry docstrings across orchestrator, node-supervisor, and ingestion-worker no longer reference a specific model
- fixSession mock data no longer returns a hardcoded model name
v1.4.9
2026-03-21- fixInstaller now exits cleanly after completion instead of hanging until Ctrl+C
v1.4.8
2026-03-21- fixLicense validation endpoint now works in production — added missing Supabase env vars to Vercel deployment
- fixStripe production key was using test mode — replaced with live key
- fixCreated `licenses` table in Supabase with proper grants (table had never been applied)
- fixStale pricing in schema.org JSON-LD ($29 → $99) and waitlist email ($49 → $99)
- docsExpanded documentation with 7 new pages: FAQ, Licensing, Troubleshooting, Configuration, Starter Agents, Pro Agents, Olivia orchestrator
v1.4.7
2026-03-21- featA.L.I.C.E. | Control installs now carry `teamId` through runtime node registration, heartbeats, and hosted session event delivery
- featCloud config persists team identity metadata so upgraded installs can attach themselves to the shared hosted control plane automatically
- fixHosted admin and Mission Control telemetry now line up with team-scoped Supabase data instead of relying on local-only runtime assumptions
v1.4.6
2026-03-21- featInstaller now ships a bundled `coding-agent` skill that prefers Codex when OpenAI is the default runtime model and Claude Code when Anthropic is the default
- featPublished package now includes the real Starter and Pro agent registries, with Pro installs assembled from the Starter roster plus the 18 Pro add-ons
- fixPro license handling now stores structured validation state, supports a short explicit grace window during service outages, and reports provisional status clearly in `--doctor`
- fixInstall state is now more stable across updates: manifests retain skills and package version, `agentToAgent.allow` preserves existing non-A.L.I.C.E. entries, and `--doctor` checks the expected tier roster more accurately
- docsCLI help and docs now match the supported automation flags for `--tier`, `--license-key`, and `--coding-tool`
v1.4.5
2026-03-20- featA.L.I.C.E. | Control nodes now report heartbeats to `admin.av3.ai` as well as the hosted runtime, making the admin portal's Fleet and Incidents views truly node-driven
- featCloud installer config now records the admin heartbeat endpoint alongside dashboard, ingest, and runtime URLs
- featShared worker-token auth now supports machine-to-admin heartbeat delivery for hosted Mission Control nodes
v1.4.4
2026-03-19- featPro installs can now enable the A.L.I.C.E. | Control add-on directly from `npx @robbiesrobotics/alice-agents`
- featInstaller now bundles and installs the `mission-control-bridge` plugin, then enables it in `openclaw.json` automatically
- featMission Control cloud settings are now written to a portable local config so fresh installs work across macOS, Linux, and Windows without hard-coded paths
- docsREADME and CLI help now document the new `--cloud`, `--cloud-token`, and Mission Control setup flow
v1.4.3
2026-03-19- featOlivia now presents herself as **A.L.I.C.E.** by default while still responding to `Alice` and `Olivia` as valid aliases in OpenClaw/NemoClaw sessions
- featOrchestrator templates now sign off as **A.L.I.C.E.**, making the installed workspace identity consistent across prompts, memory, and user-facing replies
- docsPackage guidance now reflects the new orchestrator naming and release workflow for future npm + landing-site updates
v1.4.2
2026-03-19- featA.L.I.C.E. agents are now model-agnostic by default and inherit the configured OpenClaw/NemoClaw runtime default
- fixOlivia is no longer pinned to a specific provider or model during install, avoiding accidental overrides of the user's existing runtime defaults
- docsPackage docs now clarify that auto-detect uses the runtime default model/provider rather than assigning a hard-coded model to the agent team
v1.4.1
2026-03-19- fixInstaller now guards against provider/model mismatches before rewriting `agents.defaults.model` or Olivia's default model
- fixOpenAI preset now falls back to your existing detected OpenClaw model when the required provider is not actually configured
- featCompatibility checker now flags model/provider mismatches in `openclaw.json` before they become runtime breakage
- fixInstall manifest now records the effective model preset that was actually applied, making support and debugging much clearer
- docsNon-interactive `--yes` install behavior clarified across package and site docs
v1.4.0
2026-03-19- featCross-platform installer — Windows, macOS, and Linux all handled correctly with platform-specific install paths
- featWSL2 wizard for Windows — installer detects WSL2 state and walks users through setup: installs Ubuntu, sets WSL2 as default, then runs NemoClaw inside WSL2
- featNemoClaw is now the default runtime on Linux (recommended over bare OpenClaw)
- featmacOS path — OpenClaw installed cleanly with a clear note that NemoClaw requires Linux; Brev remote deploy mentioned as the path to sandbox mode
- featSkills & Tools system — new `--skills` command to browse, install, and remove agent skills via `clawhub`
- featSkills wizard step baked into install flow — 5 curated categories: Productivity, Web & Research, Communication, Developer Tools, Smart Home & IoT
- featNemoClaw policy auto-apply — when NemoClaw is the runtime, skill installs automatically add required egress endpoints via `openshell policy set`
- featBrand terminal overhaul — green `#76b900` ANSI color scheme, `printBox`, `printSection`, icon set (✔ ✗ ⚠ ℹ), NO_COLOR/CI-safe via `colors.mjs`
- feat`--doctor` now checks installed skills are present on disk
- fix`which` replaced with cross-platform `commandExists()` — Windows no longer crashes on install
- fix`process.env.HOME` replaced with `homedir()` everywhere — correct on all platforms
- fixWSL `--list --quiet` decoded as UTF-16LE — distro detection now works correctly on Windows
- fix`wsl --set-default-version 2` enforced after WSL install — agents always get WSL2, never WSL1
v1.3.2
2026-03-17- fixVersion bump to align with npm release
v1.3.1
2026-03-17- featPro CLI license gate — `npx @robbiesrobotics/alice-agents` now prompts for your `ALICE-XXXX-XXXX-XXXX` key when selecting Pro tier
- featLicense validation — key checked against `getalice.av3.ai/api/license/validate`; fails open (offline-safe) so install works without internet
- featLicense stored at `~/.alice/license` — enter once, remembered forever
- feat`--doctor` now shows license status — Pro license key (masked), or Starter tier confirmation
- feat`--yes` flag gracefully falls back to Starter if no stored Pro license
- fixLinux Docker permission detection
- fixInvalid license key triggers 3-attempt retry with helpful error messages before graceful Starter fallback
- fixPricing copy updated across all docs — Free / Pro $49 one-time / Cloud Add-on $20/mo (requires Pro)
v1.3.0
2026-03-17- feat`--doctor` flag — full health check: runtime, config validity, agents present, workspaces on disk, model configured
- feat`--update` flag — one-command non-interactive upgrade (`npx @robbiesrobotics/alice-agents --update`)
- feat`--version` / `-v` flag — print installed package version
- featMEMORY.md scaffolded into every new agent workspace from day one
- feat14-test suite added using Node built-in `node:test` (zero external deps)
- ciGitHub Actions CI for npm package: syntax check across Node 18/20/22
- ciGitHub Actions auto-deploy for landing site (fixes broken Vercel git integration)
- docsGetting-started docs updated with Model Setup section; Prerequisites no longer implies a specific API key is required
- fixREADME Requirements section updated — no specific API key required
v1.2.9
2026-03-17- featModel auto-detection — installer now reads your existing OpenClaw/NemoClaw config and uses your configured model by default
- fixRemoved hard-coded Anthropic/Claude defaults — no Claude API key required to install
- feat`--yes` non-interactive flag now uses your detected model instead of forcing Sonnet
- featInteractive installer shows your current model as the first (recommended) option in model selection
- fixCustom model preset no longer falls back to `claude-sonnet-4-6` when left blank
- docsREADME updated to clarify A.L.I.C.E. is model-agnostic — works with any provider configured in OpenClaw
v1.2.4
2026-03-16- featNemoClaw compatibility verified — A.L.I.C.E. agents run inside OpenShell sandbox
- featAnnual billing toggle on pricing page (save 2 months)
- featFounder pricing progress bar with live spot counter
- featBlog system with MDX support — 5 launch posts
- featContact form with Resend email delivery
- featPro agent grid consolidated — single unlock overlay
- featAgent hover tooltips with domain bios
- featFAQ reordered + 3 new questions (API keys, vs ChatGPT, production-ready)
- fix"Install free" CTA now links to install flow, not pricing
- fix"Start Pro trial" → "Start Pro" (no misleading trial language)
- fixTeam CTA "Contact us" → "Talk to us" → /contact
- securityX-Frame-Options, X-Content-Type-Options, Referrer-Policy headers added
- featsitemap.xml, robots.txt, JSON-LD structured data
v1.2.3
2026-03-15- feat28 agents — added Starter and Pro tier separation
- featSupabase auth flow for Pro signup with browser token callback
- featMission Control dashboard integration
- fixChat history persistence across sessions
- fixSupabase DAL wired — data persists across container restarts
v1.2.2
2026-03-14- featInitial 28-agent roster with SOUL/AGENTS/PLAYBOOK workspace templates
- featMulti-model support: Anthropic, OpenAI, Ollama
- featInteractive installer with Fresh/Merge/Upgrade modes
- feat--yes flag for non-interactive install
A.L.I.C.E. follows semantic versioning. Breaking changes are rare and always flagged.