Everything on getlobe.dev

What Lobe does: a local proxy that splits every request into DNS, TCP, TLS, TTFB, and download phases, judged against grounded latency baselines.

Free CLI forever; Pro adds cloud session storage, route trends, and share links; Team adds the GitHub PR bot and CI checks. Billing and refund details.

Drop a HAR file exported from Chrome DevTools and get a phase-by-phase latency breakdown in the browser — no proxy setup, no account.

Set up the GitHub PR bot: capture performance sessions in CI, grade pull requests against your main-branch baseline, and fail builds on SLA breaches defined in lobe.yml.

How Lobe handles HTTPS without certificates, measuring latency under concurrency, its own proxy overhead, what lobe explain sends to Anthropic, and whether the CLI phones home.

Plain-language explanations of HTTP latency: TTFB budgets, connection phases, pool exhaustion, and bimodal distributions.

The usual suspects when localhost feels slow: IPv6 DNS resolution, missing keep-alive, dev-mode framework overhead, cold caches, and small local DB pools — and how to tell them apart.

Realistic time-to-first-byte budgets for APIs by environment — loopback, LAN, and remote — plus why the popular 800ms web guidance doesn't apply to API endpoints.

Latency is the network round trip; TTFB includes it plus your server's thinking time. How the two relate, and which one to optimize when.

Fast for one request, slow under load, timeouts at peak: how an exhausted database connection pool behaves, and the bimodal latency signature that identifies it.

When response times cluster into a fast group and a slow group, an average lies to you. What causes two-peaked latency — cache misses, pool waits, cold starts — and how to detect it.

Copy-paste Server-Timing middleware for the major frameworks, so any client or profiler can split time-to-first-byte into db, render, and cache segments.

Three ways to break a request into connection phases — curl -w, browser DevTools, and a local profiler — and what each phase should cost.

How to attribute page latency between your own endpoints and external dependencies like Stripe or OpenAI — per-host timing, dependency budgets, and the keep-alive trap.

Install with cargo: the open-source Rust CLI for capturing and profiling HTTP request latency locally.

brew install kpwithcode/lobe/lobe — prebuilt binaries for macOS (Apple Silicon + Intel) and Linux.

Subscription terms, cancellation, refunds, and acceptable use.

What data Lobe stores, what never leaves your machine, and what lobe explain transmits.