Back to services

SegroveLens

Evidence-based software analysis for embedded and connected products

Software is becoming a larger and more critical part of physical products. Firmware and embedded software increasingly work together with gateways, backend services, applications and service tools. At the same time, development is accelerating, AI-generated code is becoming common, and requirements for security, quality and traceability are increasing.

Most product companies already use code-quality, security and dependency tools. The challenge is turning fragmented findings into a reliable view of the product’s real software risks—and deciding what to address first.

THE BASELINE

From hidden risk to a prioritised improvement plan

Segrove Lens analyses the agreed parts of the product software and creates an evidence-based baseline across areas such as:

  • Security
    • Vulnerabilities in your code: taint analysis across ten languages, mapped to CWE
    • Leaked secrets across the full git history
    • Weak crypto and hard-coded credentials
    • Supply-chain risks, malicious packages and known CVEs ranked by reachability
  • Code quality and maintainability
  • Test quality: complexity combined with coverage into a CRAP score per function
  • Architecture and structural risks
  • Build, release and delivery readiness

Segrove Lens combines structural code intelligence, deterministic checks and controlled AI-assisted analysis. Findings link to real files, symbols and lines of code; anything that cannot be confirmed is marked unverified. It works on mature codebases with years of history and on new projects alike, and the same checks apply whether the code was written by people or by coding agents.

Consistent results, run after run. Scans and metrics come from deterministic methods with fixed, versioned rules. The same code gives the same evidence, so a change in the results reflects a change in your code, not in the tool.

See your progress over time. Each new run is compared with your baseline. Every finding is tracked as new, still open or resolved, and you see per category and severity whether things are improving, stable or degrading. Issues surfaced only because a new check was added are shown separately, never counted as new problems.

UNDER THE HOOD

The analysis graph. Evidence before opinion.

Every Segrove Lens run follows the same graph, repository by repository. Deterministic tools build the facts first. AI analysts interpret them, and every AI finding must pass a verification gate before it reaches the report.

Deterministic, no AIAI analyst, verified

  1. 01Index

    Structural code index and ranked repository map. tree-sitter grammars for ~40 languages, Roslyn for C#. Git history as facts.

  2. 02Scan

    Source-to-sink taint analysis, secrets across the full git history, weak crypto, licences, dependency inventory, dead code, cycles, hotspots, CRAP.

  3. 03Analyse

    Specialist analysts for security, architecture, code quality, testability, dependency health, privacy and AI exposure.

  4. 04Verify

    Each finding cites file, line and a verbatim quote. A deterministic gate checks the quote exists there. Unconfirmed findings stay visible, marked unverified.

  5. 05Score

    Cross-repository analysis and a fitness scorecard. Ratings are computed by code; AI only writes the narrative. Steps that did not run are disclosed, never shown as clean.

  6. 06Report

    Fitness report with executive summary, per-repository deep dives and a top-10 action list. Evidence in SARIF.

TEST RISK

The CRAP score: complexity meets coverage

Coverage alone hides risk: a simple getter without tests matters little, a complex function without tests matters a lot. As part of every Segrove Lens run, a CRAP score (Change Risk Anti-Patterns) is computed for every function by combining its cyclomatic complexity with its measured test coverage. It uses the coverage your test suite already produces; where no coverage has been measured, the report says so instead of showing a clean score.

CRAP = complexity² × (1 − coverage)³ + complexity

Functions above the threshold are where a change is most likely to break something unnoticed. They become concrete, located backlog items: add tests here, or simplify this.

Computed without AI from complexity and coverage data.

High risk Test coverage → Complexity →

THE DELIVERABLE

What you receive

A Segrove Lens Baseline includes:

  • An agreed baseline for the selected repositories, product versions and software layers
  • Important findings with traceability to the code
  • A clear account of what could—and could not—be analysed or verified
  • A report and scorecard for both technical and non-technical readers
  • A joint findings and prioritisation workshop
  • A practical, prioritised backlog with recommended next steps

Your team can then make the improvements independently, work with Segrove, or use Segrove Compose.

Segrove Lens and Segrove Compose can operate on customer-controlled infrastructure with local models. With a correctly configured Offline profile, local model endpoint and controls on outbound network traffic, the complete analysis can operate with zero cloud egress.

LET'S START WITH WHAT MATTERS

Know what to fix next.

Start with a conversation about your product, your software and the decisions you need to make with confidence.

Email us

Tell us about your product, codebase or question. We reply by email.

reza@segrove.ai Write to us

We use your details only to respond to your enquiry. Privacy notice