Primary skill
Prior-art-search skill
Atlas searches first and then reads at claim level. Atlas keeps Articles, Guidelines, and Board practice aligned.
Third Generation
Atlas Platform
Patent Intelligence. Reimagined.
Skill-based autonomous agents that understand context and deliver patent-grade results.
Third Generation
Autonomous Agents That Actually Understand
Our third generation introduces a fundamentally new architecture: Skill-based Micro-Agents orchestrated by a Hybrid Intake system with deep Capabilities understanding.
Skill-Based Micro-Agents
- Not prompts — Skills
- Recursive & connected
- Executes domain logic
Hybrid Intake Algorithm
- Instant understanding
- Context-aware routing
- Adapts to your workflow
Capabilities System
- Knows what you need
- File intelligence
- Feels like tailwind
Traditional AI agents either hallucinate (too autonomous) or feel rigid (too scripted). The third generation Skill-based architecture gives you the best of both: autonomous agents that are precise, flexible workflows that adapt to your day, and automated processes that actually understand patent work.
How Third Generation Atlas Thinks
Watch how our autonomous agents plan, execute, and validate — step by step.
01
InputYour question, document, or task
02
Hybrid IntakeFast understanding of what you need
03
Capabilities RouteMatches intent to agent workflow
04
Skill Orchestration
Skill A
Skill B
Skill C
Execute
Micro-agents execute: skills → skills → action
05
ValidateAutomated quality checks
06
ExplainFull reasoning trail
07
OutputYour answer with sources
Atlas, unpacked
Why Atlas feels better in practice
The internal explainer ideas now live here in the site’s visual language: as interactive views into the stages, surfaces, and system layers.
03 / Execute Loop
Pulls real sources into execution.
Reader, retrieval, and legal sources keep the answer tied to patents, documents, and law.
ReaderLegal retrievalDocument grounding
Tools before claims.
Documents stay in the loop.
Why this matters
This is where Atlas earns trust: sources come before prose.
Third Generation Architecture
Skills, Not Prompts
Forget prompt engineering. Third generation Atlas runs on Skills — self-contained autonomous units that think, reference, and execute.
Traditional Prompts
- ✕Static text templates
- ✕One prompt = one task
- ✕No logic execution
- ✕Breaks with edge cases
- ✕80+ prompts to manage
Third Generation Skills
- Dynamic, context-aware
- Skills reference each other
- Runs domain logic
- Adapts to any workflow
- One unified system
Recursive References
Skills call other skills intelligently
Domain Logic Execution
Not just text, but actual patent logic
Context Adaptation
Same skill, different behavior based on context
Workflow Scaling
Add new agent capabilities without breaking existing ones
Interactive Matter Simulator
Pick a patent task. See what Atlas leans on.
This workbench is inspired by the internal Atlas explainer: change the matter, jurisdiction, evidence state, and surface to see which skills, tools, and safeguards dominate.
Matter
Jurisdiction
Evidence
Surface
Prior Art Search / EPO / Live Chat
Prior Art Search
Find the right references fast without sounding certain too early. Guidelines, Articles, and Board practice. Fast loops with the attorney or team in the loop.
Live retrievalClaim-level readingFamily awarenessEPCGuidelinesBoA
Tool stack
Patent searchPatent detail viewFamily viewEPC materialsEPO GuidelinesBoA decisionsCitation check
Why Atlas wins
Atlas searches first and then reads at claim level. Atlas opens the search cleanly and actively builds the evidence set. Fast, but still grounded.
Generic AI risk
Generic models often synthesize before the evidence set is stable. Without real search, the prose gets fake quickly here.
Most relevant stage
Execute Loop
Closest surface
Live Chat
Dominant layer
Tools
Best surface
Live Conversation
For quick refinement loops and close collaboration.
Grounding98
How strongly Atlas pulls in real sources.
Legal Precision88
How forum-sensitive and precise the logic becomes.
Drafting Fit32
How naturally the output fits language work.
Research Stamina64
How much Atlas leans into longer work.
File & Context Handling That Feels Like Tailwind
The third generation Capabilities System gives Atlas a basic understanding of what patent attorneys actually want.
Smart File Understanding
Upload a patent document, office action, or prior art — our third generation agents automatically understand the document type and know what workflow you need.
Context Carries Forward
Your current case, client specifics, and recent work inform every agent response. The automated workflow remembers.
Adaptive Agent Responses
The same question triggers different agent workflows depending on context — quick check vs. deep analysis.
What Agents Can Do
Automated Workflows for Every Patent Task
Each capability is powered by specialized autonomous agents running on the Skills system.
Prior Art Agent
Automated novelty searches with intelligent query expansion
Drafting Agent
Claim drafting workflows with built-in support checking
Office Action Agent
Autonomous analysis and response drafting
FTO Agent
Automated claim scope analysis and risk mapping
PSA Agent
Problem-Solution Approach with technical effect workflows
Research Agent
Legal research across EPO decisions, UPC, and EPC
Connected Knowledge
Direct Access to Authoritative Sources
Atlas doesn't guess — it searches. Direct connections to patent databases, case law, and legal resources ensure every answer is backed by real sources.
Worldwide Patents
EPO, USPTO, WIPO, and national patent databases
EPO Boards of Appeal
Complete database of Board of Appeal decisions
UPC Case Law
Unified Patent Court decisions and orders
German BGH
Federal Court of Justice patent decisions
EPC & Guidelines
European Patent Convention and EPO Guidelines
Scientific Literature
Academic papers and technical publications
Web Search
Targeted internet searches for non-patent literature
Your Documents
Upload and analyze any patent document with OCR
Model Infrastructure
European AI. Global When You Need It.
Atlas defaults to European AI models for data sovereignty. When you need the absolute cutting edge, unlock global models with one click.
Active by default
European by Default
All your data stays in the EU. Atlas dynamically selects the best available model across our European partners — no configuration needed.
🇫🇷ScalewayFrench cloud infrastructure with GPU instances
🇩🇪StackItSchwarz Gruppe · German cloud infrastructure
Atlas automatically picks the provider with the most advanced model available.
Enable per session
Global Models, On Demand
Need the most powerful models available? Enable global mode per session to access the latest from the world's leading AI labs.
Google Gemini
OpenAI GPT
Anthropic Claude
When enabled, data may be processed outside the EU for that session. All other sessions remain European by default.
Built-In Outage Protection
Atlas is a platform, not a single provider. If one infrastructure partner goes offline, we automatically route traffic to another — no downtime, no manual intervention. With multiple providers across Europe and globally, Atlas stays operational even when individual providers don't.
2 European providers
Additional global providers
Automatic failover routing
The Brain Behind Every WunderIP Product
Atlas isn't a standalone product — it's the intelligence engine that powers the entire WunderIP suite.
When you use Wunder Chat for research or Wunder Word for drafting, you're using Atlas. The same powerful AI, the same connected data sources, the same patent-grade quality — seamlessly integrated into every tool.
Enterprise
Atlas as a Service
Want to integrate patent AI intelligence into your own systems? We offer Atlas as a standalone service for enterprises and legal tech companies.
- Full API access to Atlas capabilities
- Custom model training on your document corpus
- Dedicated European infrastructure
- White-label integration options