Pre-production observability for .NET & Rust

Catch CI regressions
on the pull request.

Stop losing developer hours to blind CI reruns and slow builds — and catch regressions before they reach production. Wrap your existing CI commands with zero changes to application code. Pulse correlates compile logs (.binlog, cargo --timings), test results (.trx, nextest), benchmarks, load tests (k6, NBomber), OpenTelemetry, and SARIF with live hardware telemetry from your CI runner. Developers get a strict delta summary on the PR — backed by a full forensic dashboard when you need root cause.

  • Save developer time — no blind reruns or raw log digging
  • Catch it before production — perf, build, and test deltas on the PR
  • Runner noise ≠ flaky test — throttled CI VMs filtered from baselines
  • Strict delta against baseline — zero legacy warning floods, zero SDKs
Codebase Pulse dashboard showing build history, artifacts and diagnostics

Why it pays back

Save developer hours. Catch it before prod.

Every blind CI rerun, slow compile, and escaped regression costs engineering time. Codebase Pulse correlates build, test, benchmark, load-test, trace, and SARIF artifacts with runner hardware telemetry — turning noisy CI output into a trusted pull-request loop.

  • Developer time saved

    Stop rerunning CI until green or digging through 100 MB logs.

    Instead of downloading .binlog or .trx files locally to guess why a job failed or slowed down, developers get the exact project, analyzer rule, test case, or benchmark diff right in the PR comment.

  • Shift-left detection

    Catch N+1 queries, memory spikes, and race conditions before merge.

    Benchmarks and OpenTelemetry traces from CI expose allocation spikes and N+1 call-count jumps; parallel test analysis catches race conditions. Fixing them takes minutes while the PR context is fresh — instead of an on-call incident in production.

  • Zero alert fatigue

    When Pulse flags a pull request, the team knows it is real.

    Throttled cloud VMs and infrastructure hiccups are automatically marked as Noisy Runs and excluded from baselines. Legacy warnings stay in the baseline so PRs only show what your branch changed.

How the signal stays clean

  1. 01

    PR-first, dashboard-backed

    One aggregated Markdown summary and optional SARIF inline annotations on the pull request — or in the CI job output when there is no PR. Open the web UI only when you need charts, history, and forensic drill-down.

  2. 02

    Strict delta & capped noise

    Compare fingerprints against your baseline branch — not the entire history of warnings. Penalty caps and automatic eviction keep legacy debt and resolved flaky tests from flooding every run.

  3. 03

    Zero-code instrumentation

    No NuGet or crate dependencies in your application. Wrap your existing CI commands — Pulse automatically enables and ingests .binlog, Cargo timings, TRX, nextest JSON, BenchmarkDotNet, Criterion, OTel, and SARIF.

Four modules · .NET & Rust

Find it on the runner — not after the deploy.

Four modules on the runner you already operate. Depth stays in the report, headlines land on the PR or in the CI output. After baselines on your chosen branch look right, enable --guard on pull requests.

  1. Build · MSBuild & Cargo

    Pinpoint the project, crate, analyzer, or proc-macro that slowed CI.

    MSBuild binlog and Cargo timings — normalized duration, artifact deltas, Roslyn rules, rustc phases, build scripts, and duplicate crate versions.

    Explore build analytics →
  2. Test · TRX & nextest

    Know if a failure is flaky, a real bug, or a starved CI runner.

    TRX, nextest, and JUnit — noisy runs excluded, same-commit flip detection, slow tests, and parallel overlap hints for race conditions.

    Explore test analytics →
  3. Performance · BDN, k6, NBomber & OTel

    Compare benchmarks, load tests, and traces down to method and scenario.

    BenchmarkDotNet and Criterion item diffs, k6/NBomber load-test KPIs and checks, plus OpenTelemetry operation regressions — a CI signal layer, not production APM.

    Explore performance analytics →
  4. Codebase · SARIF & repo health

    New SARIF on the PR without dumping legacy debt.

    Any SARIF scanner, git bloat and LFS checks, ownership gaps — delta against your baseline branch with capped scoring for hygiene vs criticals.

    Explore codebase analytics →
  • Report

    The full picture of the run, in the product.

  • PR / CI

    Headlines only — on the PR or in the job output where the agent ran.

  • Guard

    --guard on pull requests — headline regressions vs the baseline branch you configure.

Guard

Fail the job when a headline regresses.

Establish baselines on the branch you configure first. When those runs look trustworthy, enable --guard on pull request jobs. Guard fails the CI job — and the PR — on headline regressions vs that baseline: build, tests, performance, or codebase. Not on the full report.

  1. 01

    Baseline first

    Run pulse-agent on your baseline branch without --guard (or with --warn-only ) until noisy runs, flaky lists, and legacy debt sit in a baseline you trust. Then add --guard on pull request pipelines — any branch can be the baseline via --baseline.

  2. 02

    Headlines only

    Same numbers as the PR comment or the CI output. Guard is a gate on the headlines, not a second product. The report still holds the picture of the run.

  3. 03

    Any module

    Build, tests, performance, or codebase. A noisy build, a newly-slow test, a benchmark or load-test regression, a SARIF spike — if the metric regresses, the job fails. The PR does not land either.

Security

The agent stays on your runner.

pulse-agent parses artifacts locally on your runner and transmits only anonymized metric summaries and timing graphs. Source code and raw binaries never leave your infrastructure.

Security page

How it lands

Use the runner you already operate.

Wrap dotnet or cargo commands with pulse-agent build, pulse-agent test, and pulse-agent summary. Pulse automatically enables compiler and test telemetry flags — zero packages in your application. GitHub Actions and Azure Pipelines tasks are available too.

Get

Install the CLI on the runner

Linux or Windows. Once per image, or as the first step of the job. GitLab, TeamCity, Jenkins, Bamboo, and other CLI runners.

bash

$curl -fsSL https://get.codebasepulse.com | sh

Use

Wrap build and test

Set PULSE_TOKEN, wrap your existing commands, then pulse-agent summary for one PR comment. Works on TeamCity, GitLab, Jenkins, and other CLI runners.

.NET · dotnet

# Environment: PULSE_TOKEN# Pulse auto-injects /bl, /p:ReportAnalyzer=true, and --logger:trx$pulse-agent build -- dotnet build$pulse-agent test -- dotnet test$pulse-agent summary --vcs-token ${GITHUB_TOKEN}

Guided onboarding & rollout: we land the first pipeline with your team, run on your baseline branch until the numbers look right, and then enable --guard on pull requests (Guard). Use --warn-only first if you want headlines without blocking merge.

Who it is for

Shared .NET & Rust CI. Built for every role.

From the developer waiting on a pull request to the CTO paying for runner minutes and production incidents — everyone looks at the same calibrated run, at the depth they need.

Daily PR workflow

Stop debugging CI runners and rerunning pipelines until they turn green.

When a job fails or takes twice as long, you should not have to download a 100 MB .binlog or dig through raw console logs. Pulse puts the exact delta right on your pull request.

Where it saves your time

Instant separation between a real regression in your branch, a known flaky test, and a noisy cloud VM that ran out of CPU or disk I/O.

  1. 01

    Actionable PR comment, not log archaeology

    One structured Markdown summary on the PR shows compile time, test outcomes, benchmark diffs, and new SARIF issues. Inline annotations highlight the exact line you touched.

  2. 02

    Know if it is your code or the runner

    Pulse measures CPU single-core score, RAM pressure, and disk I/O during the job. If the runner was starved, the run is badged as Noisy — so you do not waste an hour chasing a phantom slowdown.

  3. 03

    Zero changes to how you write code

    No NuGet packages, crates, or custom test attributes to maintain. Run dotnet or cargo the way you already do and open the forensic web UI only when you need deep drill-down.