IC-star
AI model by ChipAgents
The chip neural engine for full-flow autonomous engineering.
IC-star is ChipAgents' autonomous chip neural engine that delivers full-flow autonomy for digital and analog semiconductor design. It enables engineering teams to move from specification to verified outcomes up to 10x faster by integrating reasoning, physics-aware models, and design ontology within a secure on-premises or private cloud environment.
Model specs
The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.
- Launched
- 2026-10-27
- Languages
- English
Category: Semiconductors & Chips
- Access
- On-premises or private cloud deployment
- Released
- 2026-10-27
- Deployment options
- On-premises or private cloud tenant
- Data retention policy
- Zero data retention
- Workflow support
- Digital and Analog chip design
- Autonomous chip neural engine for digital and analog workflows
- Full-flow autonomy from specification to verified outcome
- Up to 10x faster PPA convergence through physical prediction
- On-premises or private cloud deployment for IP security
- Zero data retention policy for customer design data
Overview
IC-star is ChipAgents' autonomous chip neural engine designed for semiconductor engineering teams. It functions as the core intelligence layer of the ChipAgents platform, integrating an IC Design Ontology, a Reasoning Core, Multiphysics-Aware Models, and a Design-Driven Model Navigator. It is delivered as a system that can be deployed on-premises or in private cloud environments to ensure IP security. IC-star enables full-flow autonomy, moving from specification analysis through to physical implementation and verified outcomes.
Reasoning and Orchestration
- Long-horizon orchestration: The Reasoning Core plans and coordinates agents, models, and engineering tools around a specific objective to execute long-horizon workflows.
- Workflow execution: The engine executes complex tasks including root-cause analysis, PPA-aware RTL closure, and analog layout generation.
- Full-flow navigation: The system navigates from specification analysis through to physical implementation.
IC Design Ontology
- Persistent context layer: The IC Design Ontology unifies design knowledge, plans, reusable skills, and customer-specific methodologies into a persistent context layer.
- Self-improving ontology: The system continuously improves through engineering feedback and test-time exploration of prior designs.
- Contextual grounding: Every decision is grounded in the organization's specific context, including IP, specs, and engineering know-how.
Model Navigation and Physics
- Cost and performance optimization: The engine co-optimizes model choice and token consumption around the total engineering outcome.
Use cases
- Engineering teams use IC-star to execute long-horizon workflows like PPA-aware RTL closure and analog layout generation.
- Designers deploy the engine on-premises to maintain full control over proprietary IP while utilizing agentic AI.
- Verification engineers use the system to automate root-cause analysis and coverage closure with tool-validated results.
- Architects utilize the IC Design Ontology to ground AI decisions in specific organizational methodologies and prior design knowledge.
Specifications
- Access
- On-premises or private cloud deployment
- Fine-tuning
- Continuously improves through engineering feedback and test-time exploration
- Released
- 2026-10-27
- Deployment options
- On-premises or private cloud tenant
- Data retention policy
- Zero data retention
- Workflow support
- Digital and Analog chip design
How it works
- Secure Deployment: The engine is deployed within the customer's own on-premises hardware or private cloud tenant to ensure IP security.
- Objective Setting: Engineers define the design objectives, constraints, and critical judgment calls for the project.
- Autonomous Execution: The Reasoning Core plans and coordinates agents and tools to execute the workflow, from spec analysis to physical implementation.
- Verification and Feedback: Every step is validated by the customer's existing EDA tools, such as simulators and compilers, to ensure verified outcomes.
FAQ
How is design IP and data kept secure?
It deploys inside your environment, on-premises or in your own cloud tenant, ensuring RTL, specs, and reports never leave your perimeter.
How do you prevent hallucinated or incorrect results?
Each step is validated by existing tools like simulators and compilers; failed checks are fed back for another attempt until verified.
Added by AntSpace from public sources (chipagents.ai).