ClaraVerse

ClaraVerse is an open-source, privacy-first AI workspace built to replace a fragmented stack of AI chat UIs, local model managers, and automation tools with a single self-hostable platform. It is licensed under AGPL-3.0, available as a Docker deployment, a standalone desktop app for Windows, macOS, and Linux, and a hosted cloud version at claraverse.app. The project reports over 50,000 downloads and approximately 3,900 GitHub stars as of mid-2026.

What It Is

ClaraVerse is a self-hosted AI platform that combines multi-agent chat (called Crew), a layered memory system, a visual drag-and-drop workflow builder, RAG knowledge bases, and 150+ third-party integrations into one application. Users can run it entirely offline with local models via Ollama or LM Studio, connect to cloud providers like OpenAI, Anthropic, and Google, or use the hosted cloud version with TEE-protected inference. The core platform is written in TypeScript (React 19 frontend) and Go 1.24 (Fiber backend), with MySQL, MongoDB, Redis, and SearXNG bundled in the Docker Compose stack.

How the Agent and Memory Architecture Works

ClaraVerse differentiates itself from simpler chat UIs through two architectural choices: Crew and a layered memory system.

• Crew is a multi-agent team feature where users assign a project brief, hire a team of agents, and work through a Kanban pipeline (Drafts → Queued → Working → Review → Done). Every card returns to a human for review before it ships.
• Memory operates in two tiers: a pinned tier where facts are always injected regardless of relevance scoring, and a recall tier where facts are retrieved by embedding similarity against the current conversation. Unused memories decay and archive automatically. All memory entries are encrypted at rest with AES-256-GCM using a per-user key derived via HKDF.
• Clara Agent (claracli) is a terminal coding agent that ships in the same repository, shares the user's account and model configuration, and can read, write, and edit files on the local machine.

Deployment Model and Platform Coverage

ClaraVerse supports three deployment paths:

• Self-hosted via Docker: one-line install script for Linux/macOS, or Docker Compose with the full backing stack (MySQL, MongoDB, Redis, SearXNG, Qdrant for RAG).
• Desktop app: an Electron-based standalone app for Windows, macOS, and Linux, available at the ClaraVerse-Desktop repository.
• Cloud: a hosted version at claraverse.app using TEE-protected inference providers (Trusted Execution Environments), where the project states data runs in encrypted memory enclaves with zero server-side storage.

Mobile apps for iOS and Android are listed as available, with encrypted cloud sync across devices.

Integrations and Model Support

The platform auto-detects Ollama and LM Studio on a 2-minute poll and imports all local models with zero configuration. Supported cloud providers include OpenAI, Anthropic (Claude), Google (Gemini), Grok, DeepSeek, and any OpenAI-compatible endpoint. The README lists 150+ built-in integrations — including Slack, GitHub, Jira, Google Sheets, Notion, Discord, Telegram, and HubSpot — shared across Chat, Workflows, Crew, and Routines without requiring MCP. Native Model Context Protocol (MCP) bridge support is also included for tool connections.

Update: v0.3.1 — One-Line Install, RAG-Ready

The latest release, v0.3.1 ("ClaraVerse v0.3.1 — One-line install RAG-ready"), was published on May 31, 2026. The v0.3 milestone (Q1 2026) introduced the Claraverse V2 open-source codebase (the current cloud version), team collaboration features, self-hosting of models in TEE inference, and iOS/Android companion apps. Earlier milestones include v0.2 (December 2025) which added on-premise deployment, TEE-supported cloud inference, multi-agent architecture, and an MCP bridge, and v0.1.2 (June 2025) which added the visual agent builder, multi-model orchestration, RAG document processing, and voice interaction (TTS/STT). The project was created in March 2025 and has been actively pushed through mid-2026.

Why It Got Attention

The project positions itself explicitly against Open WebUI, publishing a feature comparison table in its README that highlights differences in licensing (AGPL-3.0 with no user cap vs. BSD-3 with a branding clause), multi-agent orchestration, memory architecture, and workflow automation. Community posts on Reddit's LocalLLaMA, LinkedIn, and YouTube have described it as a potential alternative to OpenWebUI, and the project's own site attributes the quote "ClaraVerse: The Open-Source AI Powerhouse That Might Just Dethrone OpenWebUI" to a dev.to blog post by Deep Shah.