Synopsys has launched AgentEngineer, a portfolio of domain-specific AI agents built on its new Autopilot Platform, and confirmed general availability for the end of 2026. More than 50 customer engagements are already underway, the company said.

The portfolio spans six named domains: verification, system validation, implementation, analog and mixed-signal (AMS) design, manufacturing, and simulation and analysis. The launch follows July demonstrations at the Design Automation Conference (DAC), where Synopsys showed agentic AI workflows developed with Nvidia and Microsoft.

Synopsys claims up to 50x faster verification closure and 20% higher coverage. The company told Tom's Hardware Premium those two figures come from its July work with Nvidia, measured against its own verification workflows without AgentEngineer. The release also cites a 30% productivity boost — the top of the 10% to 30% range Fujitsu reported for RTL code generation — plus 2x better token efficiency and lower latency, for which no figure exists.

"Compared to what the human experts would have done otherwise or are doing today," is how Anand Thiruvengadam, executive director of product management at Synopsys, framed the productivity measurements. The token-efficiency figure is customer-reported: an unnamed customer benchmarked Synopsys' agents against its own agents built on commercial agentic harnesses.

What does the platform actually do?

The portfolio covers concrete engineering tasks across the chip flow:

• Verification: spec interpretation through coverage closure, using emulation, simulation and debug
• Implementation: floorplanning, placement, routing, congestion, DFT, and timing, power and design-rule closure through signoff, from RTL to GDS
• AMS: analog design, layout synthesis, IP node migration, physical verification, transistor-level timing and characterization
• Manufacturing: process and device simulation, mask synthesis and mask data preparation
• Meshing and simulation: FEA coding for structural analysis, CFD coding for gas turbine combustion workflows, and EM coding for PCB EMI/EMC analysis
• Customer agents: agents customers build or bring, running alongside Synopsys' own

Thiruvengadam detailed three architectural layers in a launch blog post. Long-horizon AgentEngineers are "domain-specific super agents that orchestrate task agents." Task agents "complete specific, bounded engineering tasks," orchestrated by an AgentEngineer or invoked directly by an engineer. The tool layer's engines "execute the requested work" but set no goals and make no decisions.

Unlike long-running agents that may perform one activity for hours or days, long-horizon agents pursue objectives requiring hundreds or thousands of reasoning steps.

How do the agents verify their work?

The verification loop works as follows: the agent plans, orchestrates task agents, checks results, and adjusts whenever an intermediate result falls short. When a test finds a bug, a root-cause analysis agent reads logs, clusters errors, forms a hypothesis and inspects waveforms to confirm it. The agent then makes "local rewrites of the RTL to prove that the bugs have indeed been fixed" and produces a bug fix manifest. When results drift from the objective, the agent will "course correct, adapt, react," Thiruvengadam said.

Customers can choose commercial, open-source or fine-tuned language models, and deploy on Synopsys Cloud, their own cloud, or on-premises. "Bring their own LLMs and data and infrastructure," Thiruvengadam said. Access controls, encryption and runtime guardrails protect customer, partner and Synopsys IP when third-party agents share the workflow.

How autonomous are the agents?

Synopsys characterizes the agents as L5 on its own autonomy framework, introduced last year. "The original vision of L5 was fully autonomous execution. But not just fully autonomous execution, but also complexity," Thiruvengadam said, describing L5 as executing a complex workflow autonomously within human guardrails. "That was the idea, and that's exactly where we are." Cadence claimed Level 5 on its own scale at Computex in June.

Human approval checkpoints remain mandatory. "The guardrails are still going to be defined by the humans, the crucial approval checkpoints are still going to be human-driven," Thiruvengadam said. "Our customers will have to learn to trust these autonomous systems." Teams can set checkpoints where people inspect results, validate decisions and redirect the workflow, then "reduce intervention" as confidence grows.

Who is using it?

AheadComputing reported reduced manual engineering effort from RTL handoff through signoff with the Implementation AgentEngineer, without giving numbers. Its Vice President of Verification, Alon Mahl, provided the endorsement. Intel, MediaTek and Samsung endorsed the technology but offered no hard results. Synopsys' earlier AI tool, DSO.ai from 2020, has passed 100 production tape-outs; the new agents remain in engagements.

Ravi Subramanian, chief product management officer at Synopsys, said the technology lets chipmakers "accelerate their shift from AI-assisted design to autonomous engineering."

The open questions are timing and proof. Cadence expects Level 5 early access in the second half of 2026, and Siemens has promised self-verifying capabilities in forthcoming releases. Whether Synopsys reaches general availability by end of 2026 — and names a customer in production at that point — will anchor the roadmap for autonomous agents in chip design.