Features Overview

Uranus provides a comprehensive set of features for building, deploying, and managing AI agents.

What's New

Latest Features (v1.6.0 — September 2026)

• Typed Decisions - the Jev decision model routes tool categories, gates knowledge lookups, drives browser_task, judges workflow branches and closes tasks — every decision logged with its probabilities
• Tasks - per-conversation task memory with steps, waiting states and wake-ups, replacing the single ongoing-task slot
• Durable turns - turn checkpoints resume a turn after eviction; non-destructive compaction overlays replace summary rewrites
• Fused memory & knowledge - vector + keyword + exact-key recall; vector + FTS (+ AI Search) retrieval with a reranker, scoped by tenant → agent → contact
• Workflow nodes - Decision, Approval Gate, durable Await Event, typed AI Agent output
• Browser Run - Cloudflare provider by default, AX-tree snapshots, hosted Live View + handoff, recordings, browser_execute
• Platform boundary - tenants table and a product-neutral internal API; products push, Uranus never pulls

Previous (v1.4.0)

• Website Crawling - BFS crawl entire sites with delta detection, auto-import to knowledge base
• Access Control Profiles - Tag-based per-contact access to tools, data, and conversations
• Payments System - Transaction tracking, bank notification detection, x402 micropayments, Stripe integration
• Accessibility Tree Snapshots - ~100x token reduction vs screenshots for browser automation
• Vision Models - Configurable vision model with auto-fallback (Llama 4 Scout, GPT-4o, Claude, Gemini)
• MCP Server - Expose agent tools via JSON-RPC 2.0 with Bearer token auth
• Skills Marketplace - Browse and install pre-built agent skills
• Diagnostics - 7 diagnostic channels with DB + WebSocket subscribers

Core Features

AI & Conversation

Feature	Description
Chat Interface	Interactive AI chat with context management
LLM Integration	Multiple AI providers (Workers AI, OpenAI, Anthropic)
Tool Calling	Extend AI capabilities with custom tools
OpenAPI Import	Import any API spec as callable tool clusters
Code Mode	LLM writes code for complex multi-step tasks
Selective Loading	Tool categories chosen per message by the Jev router
Typed Decisions	Calibrated yes/no and choice decisions from a typed decision model, with a decision log
Tasks	Multi-turn task tracking scoped to the conversation, with wake-ups and auto-close
Vision Models	Configurable vision model with auto-fallback chain

Automation

Feature	Description
Visual Workflows	Drag-and-drop workflow builder with 30+ node types
Scheduling & Cron	Time-based automation with cron expressions
Calendar Tasks	Calendar-based task management

Browser Control

Feature	Description
Browser Sessions	Headless and headed browser automation
Human-in-the-Loop	Human takeover for complex scenarios
Browser Actions	Click, type, navigate, extract, screenshot
Accessibility Snapshots	Text-based page representation with ref-based targeting (DOM walk or Chrome's AX tree)
browser_task	One call for a multi-step goal, with a confirmation stop before consequential actions

Data & Knowledge

Feature	Description
Data Sources	Import files, URLs, databases, APIs
Website Crawling	BFS site crawl with delta detection and background processing
Vector Search	Fused vector + keyword (+ AI Search) retrieval with reranking, scoped per tenant
SQL Interface	Direct database access with query editor

Communication

Feature	Description
Contacts	Multi-channel contact management with cross-gateway memory
Access Control	Tag-based per-contact access to tools and data
Gateways	Email, Telegram, Slack, Discord, WhatsApp, X.com
Conversations	Threaded conversation tracking across channels

Payments

Feature	Description
Payment Tracking	Transaction logging, balance checking, and payment history
x402 Micropayments	HTTP 402 payment-gated endpoints with Coinbase facilitator
Stripe Integration	Checkout sessions, virtual issuing cards, and webhooks

Integrations

Feature	Description
MCP Servers	Model Context Protocol integration
MCP Server (Expose)	Expose agent tools via JSON-RPC 2.0 MCP protocol
Services	Third-party service connections
Agent Connections	Inter-agent communication
Skills Marketplace	Browse and install pre-built agent skills

Feature Matrix

Feature	Free Tier	Paid Tier
Agents	1	Unlimited
Workflows	5	Unlimited
Schedules	10	Unlimited
Browser Sessions	100/month	10,000/month
Vector Storage	1,000 vectors	1M vectors
AI Requests	10,000/day	Unlimited

Quick Links

Getting Started

• Introduction - What is Uranus?
• Quick Start - Set up in 5 minutes
• Architecture - System design

Common Tasks

• Create a Workflow - Build your first automation
• Set Up Scheduling - Automate with cron
• Add Browser Control - Web automation

Integration

• API Reference - REST API documentation
• Browser WebSocket API - Real-time browser control
• MCP Servers - Tool integration

Architecture Highlights

┌──────────────────────────────────────────────────────────────────────┐
│                          Uranus Dashboard                            │
├─────────┬──────────┬─────────┬─────────┬──────────┬────────────────┤
│  Chat   │ Workflow │ Browser │  Data   │  Comms   │   Payments     │
│         │  Editor  │ Control │ Sources │ & Access │                │
│         │          │         │         │ Control  │                │
├─────────┴──────────┴─────────┴─────────┴──────────┴────────────────┤
│                          API Layer                                   │
│              (REST, WebSocket, MCP Server, Callable RPC)             │
├──────────────────────────────────────────────────────────────────────┤
│                       Durable Objects                                │
│          (State, WebSocket, Scheduling, Diagnostics)                 │
├─────────┬──────────┬─────────┬─────────┬──────────┬────────────────┤
│   D1    │    R2    │Vectorize│ Workers │  Queues  │   Browser      │
│Database │ Storage  │ Vectors │   AI    │          │  Rendering     │
└─────────┴──────────┴─────────┴─────────┴──────────┴────────────────┘

Technology Stack

Frontend

• React 18 - UI framework
• TypeScript - Type safety
• Vite - Build tool
• Tailwind CSS - Styling
• Radix UI - Accessible components
• React Flow - Workflow visualization

Backend

• Cloudflare Workers - Edge compute
• Durable Objects - Stateful instances
• D1 - SQLite database
• R2 - Object storage
• Vectorize - Vector database
• Workers AI - LLM inference
• Browser Rendering API - Server-side browser for crawling and markdown extraction

Integrations

• Steel Browser - Open-source browser automation with CDP control
• MCP - Model Context Protocol (client and server)
• OpenAI/Anthropic/Google - External LLMs via AI Gateway
• A2A Protocol - Agent-to-agent communication
• Stripe - Payment processing, checkout sessions, and issuing cards
• Coinbase (x402) - Micropayment facilitation