Who is this guides library for?

Engineers evaluating self-hosted error monitoring, OTLP observability, and the cost of Sentry. Each guide is useful even if you stay on Sentry, GlitchTip, Bugsink, SigNoz, or Highlight.io.

Are the guides promotional?

No. Each comparison guide ships with a "when urgentry is NOT the right pick" block. The job is to help you decide, not to sell.

How often are guides published?

Two new guides per week. Cluster hubs come online as the cluster fills out.

Guides

Long-form notes on error monitoring you can actually run.

Self-hosting, OTLP, framework error tracking, and what Sentry actually costs. Reproducible commands, public benchmarks, and an honest “where urgentry is behind” row on every comparison.

Latest

LLM Observability Explained: What to Instrument, What to Ignore, and How to Start

LLM observability splits into two layers that need different tools. The model-call layer (prompts, completions, token usage, cost per call) is owned by Langfuse, Phoenix, and Helicone. The application layer (the agent's tool calls, retries, malformed-output failures, and the exceptions that take down the request) is ordinary error tracking, and that is where urgentry sits. Four signals are worth instrumenting: tool-call errors, plan or retry loops, malformed structured output, and budget overruns. Most token-level metrics are noise once the model works. The OpenTelemetry GenAI semantic conventions are still experimental, so urgentry treats its attribute schema and alert thresholds as guidance rather than fixed constants. MCP servers are a specific blind spot worth instrumenting on their own.

AI agents & MCP observability Jun 26, 2026 13 min read

Self-hosting & ops 16

Running an error tracker yourself: sizing, single binary vs Docker Compose, hardening, backup, upgrades, and failure modes.
Alternatives & comparisons 11

Head-to-head comparisons against Sentry self-hosted, GlitchTip, Bugsink, SigNoz, Highlight.io, Rollbar, Bugsnag, and PostHog.
Evaluation playbooks 8

Decision and validation playbooks for the DSN-swap proof: compatibility checks, side-by-side runs, pre-cutover test plans.
Language & framework error tracking 15

Per-stack error monitoring with the Sentry SDK and OTLP: Go, Next.js, Django, Rails, FastAPI, .NET, Node, Laravel, Rust, mobile, and games.
Observability fundamentals 11

OTLP, sampling, profiling, session replay, native crash symbolication, and how errors fit into a broader telemetry stack.
AI agents & MCP observability 10

Self-hosted observability for AI agents and MCP servers: spans, tool calls, token cost, and degradation tracking with OpenTelemetry.
Cost 8

Methodology for modeling your Sentry bill, the real cost of self-hosting, and where the break-even sits.
Fundamentals 5

Short, definitional pieces optimized for citation: what error tracking is, what a DSN is, what compatibility-first means.