Testbench scaffolding
Agent, sequence, and driver skeletons aligned to the target interface, with optional coverage stubs.
What ChipPilot exposes publicly, and what stays gated.
ChipPilot is a repo-aware verification agent for UVM, SVA, formal, debug, and regression workflows. This public overview explains the operating model and evaluation path without publishing operator manuals, configuration schemas, agent internals, or customer deployment playbooks.
Operational manuals, environment setup, policy configuration, and customer run procedures are available to approved early-access teams under the appropriate evaluation agreement.
Evaluation Quickstart
Public examples should be enough to understand the product shape, not enough to reproduce the operating playbook. This representative intent card shows the interaction model while keeping implementation details gated.
IntentGenerate an SVA property for FIFO overflow,
run approved gates, and return evidence.
run approved gates, and return evidence.
In a real pilot, ChipPilot runs inside a controlled workspace, retrieves scoped repo context, proposes artifacts, executes allowed tools, and returns an evidence report for review.
Verification Workflows
ChipPilot is organized around closed-loop verification outcomes, not a generic chat surface. Public workflow descriptions stay capability-level; customer-specific execution details are shared during onboarding.
Property authoring & proof
Assertion drafting with bounded proof, counterexample triage, and failure summaries scoped to the run.
Failure → patch loop
Waveform-aware root-cause hypotheses, or small controlled patch output with review context.
Triage & bucketing
Bucketed regression failures with reproducible commands and links back to the originating intents.
Evidence Artifacts
Every run produces a structured evidence packet. The shape is public; the contents of any real run remain inside the customer environment.
Generated collateral
UVM source, SVA properties, or small controlled patch output with review context.
Tool outcomes
Simulator pass/fail, bounded proof, and failure summaries scoped to the run.
Repo provenance
File/line citations into the customer repo that justify generated artifacts and repair decisions.
Run transcript
Tool invocations, write paths, command approvals, and run status — all retained for review.
Security Model
ChipPilot is designed for protected silicon IP. Public documentation should make the posture clear while keeping exact deployment and policy runbooks gated.
Runs where IP lives
On-prem, private cloud, VPC, and hybrid deployment options for enterprise evaluation.
Sandboxed execution
Command allowlists, write-path restrictions, diff budgets, secret redaction, and audit logs.
No-retention modes
Customer-controlled routing policies for private models, hosted models, or hybrid operation.
Enterprise Pilot Path
The right public call-to-action is an evaluation flow, not a complete self-serve implementation manual.
-
1
Architecture review. Confirm repo shape, toolchain boundaries, deployment target, and security constraints.
-
2
Sandboxed dry run. Execute against a scoped design slice with read/write policy and tool gates agreed upfront.
-
3
Evidence review. Inspect generated collateral, transcripts, citations, diffs, and audit trail with DV/formal owners.
-
4
Deployment plan. Decide on on-prem, private cloud, or hybrid rollout and define success gates for the next phase.
Private Documentation
Approved teams receive the operational docs needed to run ChipPilot safely without exposing implementation details on the public site.
Operator reference
Approved usage patterns, environment options, output modes, and troubleshooting guidance.
EDA setup guides
Simulator, formal, CI, artifact store, and policy configuration for approved environments.
Deployment operations
Network, secrets, RBAC, no-retention, audit export, upgrade, and incident workflows.
Need the full CLI and deployment docs?
Request early access and we will share the detailed documentation in the right evaluation context.