Product

One copilot for the whole
RTL correctness workflow

Start by reviewing and verifying the RTL you already have. Grow into generating new blocks that are correct by construction — and signed off automatically.

01 · Review

RTL review & lint

The lowest-trust-barrier entry point. Nexeda runs static lint (Verilator) plus structural checks, then an AI reviewer reads your design for functional and synthesizability risks and writes assertions to lock down invariants.

  • Categorized findings with severity & cited signals
  • Suggested SVA you can paste straight in
  • Never rewrites silicon you didn't ask it to
Functional Lint CDC Assertions

Pure critique. You stay in control of every change — ideal for sign-off review, onboarding legacy IP, or a second pair of eyes on a tricky FSM.

02 · Verify

Self-checking verification

Give Nexeda a design and a spec; it builds a SystemVerilog testbench with directed corners, a constrained-random phase, assertions, and a reference model — then runs it and scores the result.

  • PASS/FAIL per check + coverage summary
  • Waveform (VCD) captured for every run
  • Watchdog + safety scan on every simulation
PASS: reset_empty PASS: full_after_DEPTH_writes PASS: fifo_order_drain PASS: count_steady_on_simul_rw NEXEDA_SUMMARY passed=34 total=34
03 · Generate

Spec → synthesizable RTL

For a curated library of standard FPGA blocks — FIFOs, arbiters, AXI-Lite/APB slaves, CDC, FSMs — describe what you need and get production-quality Verilog that's been verified end-to-end before you ever see it.

  • No latches, explicit reset, parameterized
  • Human-in-the-loop: you approve before it ships
  • Every iteration kept as an inspectable artifact
FIFO Arbiter AXI-Lite APB CDC sync FSM

A bounded, well-understood library is where the loop converges reliably today — and where the dataset that unlocks broader generation comes from.

04 · PPA

PPA-aware, FPGA-first

On every sign-off, Nexeda runs synthesis (yosys, FPGA flow) and reports LUT/FF/cell utilization and an estimated Fmax — and biases generation toward area or timing based on your end application. Full Vivado place-and-route with timing-driven iteration is on the roadmap.

  • Utilization & Fmax estimate per run
  • Area-vs-timing bias from your spec
  • Roadmap: Vivado P&R, timing closure, ASIC flow
ppa sync_fifo (xc7a) LUTs ........ 78 FFs ......... 140 est. Fmax ... ~210 MHz status ...... SIGN-OFF ✓
Trust & security

Your RTL is your crown jewel

Nexeda is designed to run where your IP already lives. A self-hosted / VPC deployment is a first-class path, not an afterthought.

🏢

On-prem / VPC

Run the full loop inside your own environment — no RTL leaves your network.

🧰

Open toolchain

Built on Icarus, Verilator and yosys, so the verification you depend on is inspectable and reproducible.

🗂️

Every artifact saved

RTL versions, testbenches, sim logs, waveforms and PPA reports — auditable at each iteration.

See it run on your kind of block

Try the loop on a FIFO, arbiter or FSM in the live Playground.

Open Playground How the loop works