Helix AI Labs Privacy. Control. Observability. Transparency.

Valence

Downloads & Release Notes

Latest build, requirements, installation notes, and changelog.

Current release: v1.4.5 Early Access, updated 2026-03-24. Download the MSI, confirm requirements, work through install notes, or jump straight to the release history.

Download Installer (.msi) Jump to Release Notes

Latest download

Valence v1.4.5 — Early Access

Updated 2026-03-24

A unified AI chat interface with multi-provider support. Switch between GPT, Claude, Gemini, Grok, and local models in one window. Now with Kids Mode and parental controls.

Download Installer (.msi) Get Valence — $20
Windows 10/11
Supported OS
64-bit
Architecture
~250 MB
Installer size
Self-contained
Runtime model

Release channel: Early Access · Last updated: 2026-03-24

System requirements

System requirements

Utility facts only: operating system, memory, storage, bundled dependencies, and trial details.

Requirements

  • Windows 10 (build 19041+) or Windows 11
  • 64-bit processor
  • 4 GB RAM minimum
  • 500 MB disk space

Included

  • Self-contained — no .NET install needed
  • WebView2 Runtime auto-installs if missing
  • 7-day free trial, no credit card required
  • All features unlocked during trial

Installation

Installation

Three steps to get started, plus the SmartScreen note and the Ollama tip that matter most during setup.

1
Download

Download Valence-1.4.5.msi using the button above.

2
Install

Double-click the MSI file and follow the installer. Accept the license agreement and choose your install location (default: C:\Program Files\Helix AI Labs\Valence).

3
Launch

Open Valence from the Start Menu or Desktop shortcut. The setup wizard will walk you through connecting your first AI provider.

Early Access Note

Windows SmartScreen may show a “Windows protected your PC” warning during install. This is normal for new software that has not yet built up download reputation with Microsoft. Click “More info”, then “Run anyway” to proceed. The installer is safe—we will add a trusted certificate as Valence grows.

Tip

Do not have an API key? Select Ollama as your provider during setup for free, fully local AI—no account or key required. Download Ollama at ollama.com.

Release notes

Release notes

Newest first. Version tags, dates, and shipping details stay visible so the page remains useful as a changelog rather than a marketing surface.

v1.4.5
Latest
March 2026

New Features

  • Native Gemini image generation — models like gemini-2.0-flash and gemini-3-pro-image-preview can now return generated images directly in chat
  • Curated model lists — provider model pickers now default to recommended models with an expandable advanced section

Improvements

  • Model capability detection now auto-checks image generation support per model
  • Model list refresh flow improved to fetch and apply current provider model metadata

Bug Fixes

  • Fixed “images not appearing in chat” by addressing request serialization, image rendering, and post-stream UI rebuild conflicts
  • Removed incorrect image generation capability flags from stale model metadata entries

Technical Changes

  • Removed redundant active-tab UI rebuild after streaming responses
  • Google API request payloads now use Utf8JsonWriter to avoid dictionary serialization edge cases
  • Updated static model definitions to current provider model sets
v1.4.4
March 2026

Bug Fixes

  • Fixed WebView2 cache preventing new features from loading after in-app updates — the browser cache now automatically clears when a new version is detected
  • Kids Mode wizard, Parent Dashboard, and other post-v1.2 features now load correctly on upgraded installations
v1.4.3
March 2026

New Features

  • Kids Mode — age-appropriate AI profiles for children ages 3–17 with five safety tiers, anti-jailbreak protections, and automatic crisis resource injection
  • Parent Dashboard — LLM-augmented conversation scanning with live progress, sentiment analysis (positive/neutral/negative/concerning), AI-generated summaries, and crisis detection
  • Parental Controls — password-protected settings lockdown, provider/model locking, daily session time limits with extension requests, and custom safety instructions
  • Kids Setup Wizard — guided 6-step setup with privacy consent, child profile creation, safety previews, icon picker, and parental controls configuration
  • Session Timer — per-profile daily time tracking with visual countdown, warning states, and parent-approved time extensions

Improvements

  • Parent Dashboard now streams scan results in real-time with a progress bar instead of blocking on load
  • Crisis detection expanded with LLM-powered analysis and comprehensive keyword fallback covering 19+ distress phrases
  • Dashboard modal widened to 900px with proper sidebar/main flex layout and scrollable conversation detail
  • Kids Mode profile persistence and settings lockdown fixes
  • Password prompt and spell check improvements for kids profiles
v1.3.0
March 2026

New Features

  • Compare Models — side-by-side model comparison: pick 2–3 providers, send the same prompt, and watch responses stream in parallel
  • Image Generation — use #image{prompt} to generate images via OpenAI (DALL-E), Google (Imagen), Azure, or xAI; images display inline in chat
  • #forget ability — selectively delete specific memories by topic or keyword from local memory
  • Window Context — press Ctrl+Alt+O and Valence automatically captures what was on your screen (screenshot or text) to give the AI instant context
  • Memory Facts — local memory now extracts structured facts from conversations for more precise recall
  • Tool Gap Banner — smart banner suggests enabling web search or other MCP tools when the AI could benefit from them

Improvements

  • Richer memory storage — memories now include extracted facts, entity tracking, and improved summarization
  • Conversation save now indexes facts into local memory for better future recall
  • Dropdown menus improved with better positioning and overflow handling
  • Tab settings now include Window Context configuration (picture and text capture toggles)
v1.2.0
March 2026

New Features

  • Scheduled Automations — run prompts on a timer with file or command input sources; results appear in dedicated tabs with optional alert-only mode
  • Command input source — automations can execute shell commands (PowerShell, cmd, Python, Bash) and feed the output to the AI
  • Test Command button — verify a command works directly from the automation editor before saving
  • Loop ability — AI can request follow-up iterations within a conversation for multi-step reasoning
  • MCP Gallery — browse and install popular MCP servers from a curated catalog
  • Node.js auto-detection — MCP settings now detect and display Node.js availability

Improvements

  • Shell detection loading state — dropdown shows “Detecting shells…” instead of appearing empty
  • Input source badges on automation cards — see “powershell”, “File”, or “MCP” at a glance
  • Run Now spinner — play button shows a loading indicator and disables to prevent double-clicks
  • Fixed shell edit race condition — editing a Command automation now reliably restores the saved shell selection
v1.1.6
March 2026

New Features

  • Automatic update check — Valence now checks for new versions on startup and shows a non-blocking banner when an update is available
  • One-click update — click “Update now” to download and install the latest version without leaving the app

Bug Fixes

  • Fixed first-run wizard freezing on the final page — the settings save could deadlock the UI thread on some systems
  • Fixed window dragging not working when native drag regions are unavailable — the JS fallback was targeting the wrong CSS selector
v1.1.5
March 2026

New Features

  • Profile slash commands — type /ProfileName message to switch profile and send in one step
  • Profile quick switcher — press Ctrl+P to search and switch profiles instantly
  • Import conversation picker — select which conversations to import instead of importing everything

Improvements

  • ChatGPT import now captures folder and project metadata
  • Bundled WebView2 bootstrapper — automatic one-click runtime installation on machines missing WebView2
v1.1.1
March 2026

New Features

  • Local Memory System — persistent memory across conversations with LLM-powered summarization and tiered recall
  • ChatGPT Import — import all your ChatGPT conversations via browser login or ZIP export
  • Claude Import — import conversations from Claude via browser login
  • Gemini Import — import conversations from Google Gemini via browser login
  • Memory Browser — search, browse, and manage saved memories from Settings

Improvements

  • Memory list now paginates (20 at a time) for better performance with large collections
  • Instant client-side filtering when typing in the memory search box
  • Import buttons now include descriptions explaining each provider’s import experience
v1.0.1
March 2026

New Features

  • Bundled Web Search — built-in MCP server for web search and page reading, no Node.js or API keys required
  • Auto-configured on first launch — search tools are ready to use out of the box
  • Live status indicators during tool execution (“Searching the web…”, “Reading page…”)

Improvements

  • Smarter intent detection — AI now reliably invokes web search when you ask it to look something up
  • MCP tool arguments are now parsed correctly (fixes “No results” from search)
  • MCP tools are always disclosed to the AI, so it knows what tools are available
  • Quick Start example in MCP settings for easy manual configuration

Bug Fixes

  • Fixed deadlock in MCP stdio transport that could freeze the UI
  • Fixed token usage timer threading crash
  • Fixed “Check for Updates” button overlapping version number in Settings
  • Fixed MCP server path resolution for bundled executables
v1.0.0
March 2026

Features

  • Multi-provider AI chat: OpenAI, Anthropic, Google, xAI, Azure, and Ollama
  • Tabbed conversations with cross-tab context sharing
  • Conversation branching — explore alternate responses without losing history
  • User profiles with custom system instructions per profile
  • Auto-triggers for quick provider switching (@Claude, @GPT, etc.)
  • MCP (Model Context Protocol) integration for extensible tool use
  • Built-in abilities: #time, #calculate
  • Voice input with push-to-talk
  • Prompt history and template system
  • Chat history with search and auto-save
  • Customizable themes and font settings
  • Global hotkey: Ctrl+Alt+O to show or hide from anywhere
  • System tray with floating restore icon

Security

  • API keys encrypted locally with Windows DPAPI — never sent to Helix servers
  • RSA-signed license keys for offline validation
  • All AI traffic goes directly to your chosen provider

Ready to get started?

Ready to get started?

One interface for every major AI provider. No middleman. No lock-in.

Get Valence — $20 Explore Features