ELECTRONICS·INSIDER
All stories
Design & EDA

Agentrys claims its Studio platform lets chip design teams assemble a custom multi-agent flow in 30 minutes

Agentrys says its Studio platform can stand up a custom multi-agent chip design flow in about 30 minutes, with a live demonstration planned for Hot Chips 2026 on 23-25 August.

Generated image

Agentrys announced on 14 August 2026 that it will demonstrate custom-built agentic design flows at Hot Chips 2026, held at Memorial Auditorium, Stanford, from 23-25 August, where the company is a Silver Sponsor[1]. The central claim: using the Agentrys Studio platform, an initial multi-agent chip design flow can be built in about 30 minutes[1].

What the Studio platform does

Agentrys positions its technology under the banner of Agentic Design Automation (ADA), which it describes as a step beyond conventional EDA - moving from automating engineering algorithms to automating engineering tasks themselves. Agentrys Studio includes hierarchical multi-agent orchestration, governed iteration loops to ensure convergence, detailed management of design objectives, and a continuously curated design knowledge base that every agent draws on, so each completed design sharpens the next.

The flow learns from available data and methods and gets better with each run. The platform is designed to run on a team's existing infrastructure and existing EDA toolchains. Production flows that mix Synopsys, Cadence, Siemens EDA, and internal tools and scripts are supported; Agentrys orchestrates across all of them, flexibly for open-ended work and reliably for structured, long-running, multi-step workflows.

Agentrys has already claimed significant improvements in areas including design verification, formal verification, RTL coding, PDK validation, waveform analysis, analog layout, and full spec-to-GDSII implementation. These are stated claims; independent verification of the results has not been published.

The problem it is trying to solve

Agentic design flows are becoming popular for many advanced chip design projects, but a challenge of implementing this technology is finding the right flow that fits the needs of a particular project from the array of pre-implemented solutions. The company's argument is that forcing teams to pick from a catalogue of fixed flows is the wrong model - each team's methodology, toolchain, and data are different enough that a purpose-built flow is necessary.

Every new AI workload wants its own datapath, its own numeric format, and its own energy-per-inference sweet spot, creating a demand curve for new chips that is vertical while the supply of expert chip design teams is flat - and that is the problem Agentrys is trying to solve.

As ADA matures, the role of designers is expected to shift from manually executing every design task to developing, supervising, and improving the agents that design chips - in this future, agents design chips, and designers develop the agents. Mark Ren, founder and CEO of Agentrys, is an IEEE Fellow with 26 years of R&D experience across IBM Research and NVIDIA Research.

What to watch

Beyond the Hot Chips exhibit table, Agentrys is presenting a SemiWiki webinar titled Agents Assemble: Build Your Chip Design Multi-Agent System in 30 Minutes, scheduled for 15 September 2026 at 10 AM PDT, in collaboration with SemiWiki.com. That session is the more detailed technical event to track: it is where the company will have to show, step by step, whether the 30-minute assembly claim holds under scrutiny from a design-engineering audience. The company has previously published a white paper describing a production-grade chip called AgentCore - a compact, 32-bit embedded-class processor built on an open, royalty-free instruction-set architecture - as evidence that the multi-agent approach can reach sign-off quality. Whether the September webinar extends that evidence to a live, audience-driven build will be the real test of the platform's accessibility claim.

Written by Electronics Insider's automated desk from the sources above and published automatically. How we work.

Related