alloomi

mcp
Security Audit
Warn
Health Warn
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Warn
  • process.env — Environment variable access in apps/web/app/(auth)/actions.ts
  • network request — Outbound network request in apps/web/app/(auth)/actions.ts
  • network request — Outbound network request in apps/web/app/(auth)/api/discord/guilds/route.ts
  • process.env — Environment variable access in apps/web/app/(auth)/api/discord/oauth/exchange/route.ts
  • network request — Outbound network request in apps/web/app/(auth)/api/discord/oauth/exchange/route.ts
  • process.env — Environment variable access in apps/web/app/(auth)/api/discord/oauth/start/route.ts
  • network request — Outbound network request in apps/web/app/(auth)/api/discord/oauth/start/route.ts
  • network request — Outbound network request in apps/web/app/(auth)/api/integrations/[accountId]/route.ts
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
Alloomi is an AI-driven workspace that monitors business communications (like Slack and email), orchestrates tasks autonomously, and acts as a proactive digital assistant.

Security Assessment
The overall risk is rated as Medium. The tool does not execute raw shell commands or contain hardcoded secrets, and it requests no explicitly dangerous permissions. However, it routinely makes outbound network requests to external services and relies heavily on environment variables to manage sensitive data. These requests are primarily located in its authentication and Discord integration routes, which involve exchanging OAuth tokens. Given the tool's core purpose, it inherently handles highly sensitive user data such as emails, calendar events, and chat logs. You must carefully review how you configure your environment variables to ensure secrets and user data are secured.

Quality Assessment
The project is actively maintained, with repository activity as recent as today. It uses the permissive and standard Apache-2.0 license. However, it currently has very low community visibility with only 5 GitHub stars. This means it has not been broadly tested or vetted by the open-source community, which lowers the general trust level.

Verdict
Use with caution: the codebase is active and properly licensed, but its early-stage maturity and deep access to sensitive communication data warrant a thorough internal security review before deploying.
SUMMARY

Alloomi is your proactive AI workspace as your work partner.

README.md
Alloomi Logo

Node.js Version Tauri Platform License Discord X

Proactive AI workspace — understands your intent, orchestrates execution, and gets things done.

Download

Alloomi is a proactive AI workspace that monitors business signals, orchestrates tasks autonomously, and tracks and validates results end-to-end. Unlike traditional AI assistants that are passive workflow tools, Alloomi acts as a proactive AI workspace that watches, learns, remembers, and acts on your behalf.

This is the open-source core of Alloomi. It includes the core infrastructure and modules, but requires you to configure your own LLM API Key, authentication, authorization, AI MCPs and skills. For the full ready-to-use product with all features enabled, please download from the official website: alloomi.ai

Features |  Difference |  Use Cases |  Installation |  Documentation |  Developing

Features

Core Capabilities

  • 📡 Proactive Awareness — monitors signals across Slack, Email, Calendar, Documents and alerts you proactively before issues escalate
  • 🧠 Long-Term Memory (Context Atlas) — persistent knowledge graphs of people, projects, and decisions; remembers context even months later
  • 🎯 95% Noise Filtering — hundreds of daily messages refined into one focused panel; tells you what you should act on
  • ⚡ Autonomous Execution — drafts replies, schedules meetings, generates reports, tracks and validates results end-to-end; supports scheduled tasks (cron-like recurring jobs) and proactively triggered tasks (event-driven actions based on signals from Slack, Email, Calendar, etc.)
  • 💬 Natural Chat — assign tasks in plain language; no complex commands to learn with powerful skills and MCP tools

Built-in File Preview

  • Documents — Docx, DOC, ODT, RTF
  • Spreadsheets — Xlsx, XLS, CSV, ODS
  • Presentations — PPTx, PPT, ODP
  • PDF — PDF files with full rendering
  • Images — JPG, PNG, GIF, SVG, WebP, BMP
  • Code — Syntax-highlighted preview for JS, TS, Python, Go, Rust, and 20+ languages and HTML preview with live rendering

Multi-Platform Access

  • Messaging Apps — Telegram, WhatsApp, iMessage, QQ, Feishu, Weixin, Dingtalk integrations with message fetching, sending, file attachments, and real-time sync
  • Desktop Apps — Native apps for Windows, macOS, and Linux with keyboard shortcuts and system tray

Enterprise-Grade Security

  • AES-256 end-to-end encryption
  • Hardware-isolated processing environments (no public gateways)
  • Zero training commitments — your data never trains public AI models
  • Local-first architecture

Agent Runtime Integrations

  • Claude Code — Anthropic's coding agent (auto-inherits ANTHROPIC_API_KEY, ANTHROPIC_BASE_URL, ANTHROPIC_MODEL from environment & skills in ~/.claude/skills) (Default)
  • Codex — OpenAI's code generation agent (Coming Soon)
  • Gemmi — General-purpose AI agent (Coming Soon)
  • Pi — Inflection AI's personal agent (Coming Soon)
  • OpenClaw — Open agent protocol & ecosystem (Coming Soon)
  • Hermes Agent — NousResearch's agent framework (Coming Soon)

Hybrid Model Architecture (Coming Soon)

  • 🤖 Hybrid Model Routing — dynamically routes tasks to optimal models (Claude, GPT, Gemini, open-source) based on complexity, cost, and latency requirements
  • 🧪 Reinforcement Learning from Feedback (RLHF) & LoRA — continuously improves task execution quality through human feedback signals
  • 🔄 Multi-Agent Debate — multiple specialized agents collaborate and debate to reach higher-quality decisions
  • 📊 Outcome Validation — end-to-end result verification with automated checks and human-in-the-loop approval workflows
  • 🧬 Adaptive Personalization — learns your communication style, preferences, and workflows over time

What Makes Alloomi Different?

Most AI assistants are workflow tools—you give commands, they execute tasks, with no knowledge of
who you are. Sometimes they surprise you in ways you didn't expect. But usually, most of the time, they
frustrate you, filled with uncertainty, and accompanied by issues of context, memory, cost, and security.

Alloomi is different: it's a proactive digital partner that watches, learns, remembers, and acts on
your behalf. The difference is architectural.

When you connect your messaging platforms and integrations to Alloomi, you don't just chat with it through
Telegram, WhatsApp, and other apps—you also sync everything with your permission: raw messages, meetings,
emails, tweets, calendar events, voice calls, and any notes or ideas you've captured. All of this—including
your conversations with Alloomi itself—becomes the single source of truth for Alloomi's brain.

Behind the scenes, Alloomi runs a background agent on a continuous sync loop, actively gathering information
from all your connected sources. An agent without this loop can only respond based on stale context. With it,
every conversation—and every moment—makes Alloomi smarter and more aligned with you. When you create a custom
agent role in Alloomi to handle one-off or scheduled tasks, this brain acts as the orchestrator, dramatically
improving execution quality.

Architecture

Alloomi implements a complete "Receive → Process → Remember → Understand → Serve" loop:

Layer What It Does
📡 Receive Multi-platform, multi-modal ingestion — IM, email, documents, files, web data, voice calls
⚙️ Process Noise reduction at scale — deduplication, OCR, ASR, intent extraction, semantic clustering
🧠 Remember Persistent knowledge graph — people, projects, decisions — survives conversations and months of time
🎯 Understand Deep semantic intent, cross-modal understanding, emotional tone, contextual relevance
🚀 Serve Proactive delivery — smart summaries, auto-replies, scheduled reports, personalized alerts

Layered Memory Architecture

  • Raw information — original messages, files, transcripts
  • Information insights — extracted entities, decisions, key events, timeline, execution diff
  • Contextual memory — recent conversation state, temporary context
  • Knowledge-base memory — long-term people/projects/preferences knowledge graph

Use Cases

  • 🌍 Global Managers — Filter time zone and language noise, capture high-value opportunities 24/7
  • 🧑‍💻 Engineers & Product Teams — Team memory that never decays, auto-generate weekly reports, eliminate context rot
  • 🚀 Founders & Salespeople — Maintain hundreds of client relationships, auto follow-ups, personalized proposals at scale
  • 🧑‍💻 Engineering Teams — Automated dev reports, issue triaging, GitHub-Linear sync
  • 📱 Social Media Managers — Autopilot X (Twitter) account with approval workflow

See here for more features and use cases.

Screenshots

  • Document Previews — Docx, PPTx & Xlsx
Docx Excel
  • Website Generation
Website Generation
  • Mutliple Connectors
Connector
  • Automation & Cron Jobs
Automation
  • Library
Library
  • Skills
Skill
  • Message Apps
Message App

See alloomi.ai for more information.

Installation

For more information about installation, please check the Installation Guide on the Alloomi GitHub Releases.

Documentation

Detailed documentation is available at Alloomi Website.

Developing

Environment Setup

Download Node.js 22+, pnpm 9+, Rust Cargo 1.88+

Copy the example environment file and configure your credentials:

git clone https://github.com/melandlabs/alloomi
cd alloomi
cp apps/web/.env.example apps/web/.env

Required Variables

Set required variables in apps/web/.env

# Generate AUTH_SECRET
openssl rand -base64 32

# Generate ENCRYPTION_KEY
node -e "console.log(require('crypto').randomBytes(32).toString('base64url'))"
Variable Description
AUTH_SECRET Authentication secret (32+ chars)
ENCRYPTION_KEY AES-256 encryption key for local data

AI Configuration

Choose your AI provider (Anthropic, OpenAI, or OpenRouter):

# Anthropic Compatible API
ANTHROPIC_BASE_URL=https://api.anthropic.com
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-sonnet-4-6

# OpenAI Compatible API
LLM_BASE_URL=https://api.openai.com/v1
LLM_API_KEY=sk-...
LLM_MODEL=gpt-4o

For embeddings (RAG / Knowledge Base), an OpenAI-compatible API key is required:

OPENAI_EMBEDDINGS_API_KEY=sk-...
LLM_EMBEDDING_BASE_URL=https://api.openai.com/v1
LLM_EMBEDDING_MODEL=text-embedding-3-small

Optional Integrations

Variable Description
BRAVE_SEARCH_API_KEY Brave Search for web content
TG_BOT_TOKEN Telegram bot token
SLACK_BOT_TOKEN Slack bot token
DISCORD_BOT_TOKEN Discord bot token
TWITTER_CLIENT_ID / _SECRET Twitter OAuth
GOOGLE_CLIENT_ID / _SECRET Google OAuth
GMAIL_CLIENT_ID / _SECRET Gmail OAuth
AUTH_SMTP_* Email SMTP server

Requirements: Node.js 22+, pnpm 9+, Rust Cargo 1.88+

Install

# Install dependencies
pnpm install

# Start desktop app (requires Rust)
pnpm tauri:dev

Build & Test

pnpm tsc          # Type check
pnpm format       # Format code
pnpm lint         # Lint
pnpm lint:fix     # Fix lint issues
pnpm test         # Run tests

Reviews (0)

No results found