Aivyx-studio

Aivyx positions itself as a fundamentally different approach to AI assistants: one where you retain complete ownership of your data and intelligence. Unlike cloud-based AI tools that require sending your prompts to external servers, Aivyx operates entirely locally using Ollama for inference, with optional connections to OpenAI, Anthropic, or Gemini when additional power is needed. The platform is built around a multi-agent architecture called "The Nonagon" — nine specialist agents that collaborate on complex tasks under the coordination of a central mission director.

The primary use cases span personal productivity, software development, research and analysis, content creation, and enterprise deployment. Developers can use Aivyx for code review, debugging, and project planning; knowledge workers can leverage its persistent memory and research capabilities; and organizations can deploy it with full multi-tenancy, SSO integration, and cost governance. The target audience includes privacy-conscious individuals, developers seeking local AI alternatives, and enterprises requiring compliant, self-hosted AI infrastructure.

• Encrypted at Rest: All data is protected with ChaCha20Poly1305 authenticated encryption using HKDF-SHA256 derived keys, ensuring your information remains unreadable without your passphrase and key material is zeroed on process exit.

• GraphRAG Memory: Aivyx builds a semantic knowledge graph from conversations using subject-predicate-object triples with BFS traversal and community detection, enabling agents to retrieve, evaluate, and self-correct their context while maintaining encryption.

• Local-First Architecture: Runs entirely on your machine with Ollama for fully local inference, eliminating cloud dependencies while optionally connecting to external providers when needed.

• Voice Agent: Real-time voice interaction via WebSocket with STT, agent reasoning with tools, and TTS streaming, including barge-in interruption support for natural conversation flow.

• Vision & Documents: Multimodal capabilities including image understanding through Claude, OpenAI, and Ollama vision models, plus PDF, XLSX, and CSV extraction with image embeddings in memory.

• A2A + MCP Protocols: Industry-standard interoperability through Google A2A for agent-to-agent communication and MCP with OAuth 2.1 for agent-to-tool integration, enabling federation for cross-instance collaboration.

• DAG Orchestration: Parallel step execution with dependency graphs, reflection loops using LLM-as-judge, and human-in-the-loop approval checkpoints for complex workflow management.

• The Nonagon Team: Nine specialist agents (Coordinator, Researcher, Analyst, Coder, Reviewer, Writer, Planner, Guardian, Executor) each with distinct identities, tools, and capabilities that collaborate on decomposed missions.

• Heartbeat System: Proactive agent awareness with periodic checks for items needing attention, where quiet ticks are free and the LLM only runs when something has changed.

• 5 Channel Adapters: Cross-platform responsiveness through Discord, Slack, WhatsApp, Matrix, and Email, each following the same secure InboundChannel trait.

• Download and Install: Use the one-line installer for Linux, macOS, or Windows, or choose from Flatpak, Homebrew, Snap, and AUR packages; the built-in aivyx self-update command keeps you current.

• Complete Genesis Setup: Launch the desktop app for the graphical 3-step wizard covering identity (name and timezone), provider selection (LLM choice with auto-detected Ollama models), and security (encryption passphrase creation with strength meter).

• Launch Your Agent: Start interacting through your preferred interface — use aivyx run assistant for CLI commands, aivyx tui assistant for the full terminal UI, or aivyx voice assistant for real-time voice interaction.

• Configure Memory and Skills: Allow the agent to build its encrypted knowledge graph automatically from conversations, or manually manage memory with commands like aivyx memory list, search, and delete.

• Set Up Channels (Optional): Configure Discord, Slack, WhatsApp, Matrix, or Email adapters so your agent can respond across platforms using the same secure message handling.

• Enable Heartbeat (Optional): Configure proactive checking with aivyx config set heartbeat.interval_minutes 15 to have your agent periodically identify items needing attention without constant LLM calls.

• Explore Federation and Nexus: Connect to federated instances for cross-agent collaboration or configure your agent to publish insights to the public Aivyx Nexus feed at aivyx-nexus.com.

• Verify Audit Integrity: Periodically run aivyx audit verify to confirm the tamper-proof HMAC chain of all agent actions remains intact.

• True Data Sovereignty: Unlike cloud AI services that store prompts on external servers and claim ownership through terms of service, Aivyx runs 100% locally with ChaCha20Poly1305 encryption and a capability-based security model where you own everything.

• Zero Telemetry Architecture: No analytics, no phone-home mechanisms, and no stored passphrase — only derived keys exist in memory, with key material explicitly zeroed on process exit via the secrecy crate.

• Tamper-Proof Accountability: Every agent action is recorded in an HMAC chain where each entry signs the previous hash, creating cryptographic proof of integrity that breaks if any entry is modified.

• Self-Improving Intelligence: The system tracks delegation outcomes, extracts success patterns, builds learned weights for specialist selection, and implements planner feedback loops — your agent genuinely improves with every task.

• Enterprise-Grade Isolation: Per-tenant HKDF key derivation with isolated directory trees, RBAC with OIDC SSO, and cost governance with per-agent and per-tenant budgets enable confident production deployment.

• Universal Interoperability: First-class support for emerging industry standards (Google A2A and MCP) backed by 50+ companies ensures your agents can collaborate across platforms and access infinite tools through protocol extensibility.

• GitHub Repository: Access the open-source codebase at github.com/AivyxDev/aivyx-app to report issues, review code, and contribute improvements.

• GitHub Discussions: Participate in community conversations, ask questions, and share use cases at github.com/AivyxDev/aivyx-app/discussions with responses from maintainers and community members.

• Documentation Hub: Comprehensive getting started guide, user manual, and troubleshooting available in the public aivyx-docs repository on GitHub.

• OpenAPI Specification: Hand-authored OpenAPI 3.1.0 specification covering all 90+ endpoints and 29 tags, with integration guides for Slack, Discord, and custom bot development.

• Disclaimer and Legal: Full beta software disclaimer, privacy policy, and MIT license terms available on the website for transparency about usage risks and permissions.

• Linux: One-line install script with Flatpak, Snap, and AUR packages available; .AppImage and .deb formats for desktop application.

• macOS: One-line install with Homebrew package support and native desktop application.

• Windows: One-line install script with native desktop application support.

• CLI/TUI/Voice: All interfaces available through the same installation with commands aivyx run, aivyx tui, and aivyx voice for terminal, full terminal UI, and voice interaction respectively.