Accelerate waveform and trace analysis with Raft.

The problem: every hardware workflow ends up with waveforms, and valuable data buried in noise. The solution: Raft, a fast waveform analyzer you can query and script, not just another viewer to squint at. It works whether you're simulating RTL, tapping an FPGA with a logic analyzer, running a SystemC model, or analyzing counter-examples from formal tools.

Raft is built specifically for hardware debugging. It understands clock domains, signal timing, and waveform semantics out of the box, so you don't have to reimplement that logic yourself, and it stays fast even on large traces. No more slow, fragile, hand-rolled Python scripts to parse and search through waveforms. Spend your time fixing bugs, not maintaining tooling.

Why Raft Changes How You Debug

• Imagine SQL, grep, and awk designed specifically for Verilog/VHDL traces to solve hardware problems.
• Use Raft anywhere from interactive debugging to building a full CI performance regression pipeline.
• Reuse your Raft scripts and gain speed as your library grows.
• Handle multi-gigabyte waveforms without slowing down.
• Mine your traces to uncover hidden information buried in data noise.
• Extend your existing workflow and get a risk-free productivity boost from day one.

Raft is built for engineers and AI agents alike. Engineers write analysis scripts, AI agents extend or run them independently, and both work through the same interface. With Raft, agents can help you with root-cause analysis, performance analysis, power optimizations, and more.

Unlike AI agents, Raft is deterministic. The same question against the same trace always returns the same answer. That determinism grounds your agents in verifiable data instead of guesses, so root-cause claims come from the waveform itself. Every result is reproducible and traceable back to the exact signal and cycle that produced it, so you can check the agent's conclusion yourself instead of taking it on faith.

Make Any Agent Master RTL Debugging

Raft's MCP server is all you need to teach coding agents like Claude or Cursor how to effectively debug and analyze waveforms. No need for custom agents made for hardware. Use the tools you already work with every day.

‹ cva6_icache.png cva6_dcache.png ›

Raft transforms even the largest traces into detailed analyses, delivering more than just PASS or FAIL results. It converts traces into charts, tables, or log files, and post-processes simulation runs for reporting or integration with other tools. It visualizes every design component so you can track progress and spot optimization opportunities or regressions. To get you started quickly, we provide ready-to-use libraries for the protocols and standards you use in your design.

bandwidth_report.csv

Drag to compare

bucket	Reads	Writes	Avg. Delay (R)	Avg. Delay (W)	Bandwidth (R)	Bandwidth (W)
0-10k	128	64	4.2 cyc	3.8 cyc	1.4 GB/s	0.9 GB/s
10k-20k	256	96	5.1 cyc	4.0 cyc	2.1 GB/s	1.1 GB/s
20k-30k	64	32	3.6 cyc	3.2 cyc	0.8 GB/s	0.5 GB/s
30k-40k	192	80	4.8 cyc	4.5 cyc	1.7 GB/s	1.0 GB/s
40k-50k	144	72	4.5 cyc	4.1 cyc	1.5 GB/s	1.0 GB/s
50k-60k	96	48	3.9 cyc	3.5 cyc	1.0 GB/s	0.7 GB/s
60k-70k	208	88	5.4 cyc	4.3 cyc	2.3 GB/s	1.2 GB/s
70k-80k	80	40	3.4 cyc	3.0 cyc	0.7 GB/s	0.4 GB/s
80k-90k	176	64	4.6 cyc	3.9 cyc	1.6 GB/s	0.9 GB/s
90k-100k	112	56	4.0 cyc	3.6 cyc	1.2 GB/s	0.8 GB/s
100k-110k	224	104	5.7 cyc	4.6 cyc	2.4 GB/s	1.3 GB/s
110k-120k	72	32	3.2 cyc	2.9 cyc	0.6 GB/s	0.4 GB/s
120k-130k	160	72	4.4 cyc	3.8 cyc	1.4 GB/s	0.8 GB/s
130k-140k	136	64	4.1 cyc	3.7 cyc	1.1 GB/s	0.7 GB/s
140k-150k	192	96	4.9 cyc	4.2 cyc	1.8 GB/s	1.1 GB/s

↕ ↔