# APM for PHP landscape in 2026

Source: https://blog.fortrabbit.com/apm-for-php
Created: 2026-07-13
Author: Frank Lämmer
Tags: opinion

> A take on PHP application performance monitoring and meta-observability.


We are a PHP host. Our customers ship PHP code. When that is slow or blocking, it often ends in a 503 or 504. Then the customer shows up in support, mad about server errors.

We recently put together a section on :ContentLink{href="/integrations/apm/intro" text="APM tools" prefix="docs"} in our docs. This post is a snapshot of the scene in early 2026 and our state of thinking.

## Profiler, error tracker, monitor

Let's learn the categories:

- **Profilers** tell you where time and memory go inside a single request. Think Blackfire, Tideways, PHP SPX, Excimer.
- **Error trackers** capture exceptions and stack traces from production. Think Sentry, Flare, Bugsnag.
- **APM** sits between them — sampled traces, slow request lists, database query breakdowns, alerting. Think New Relic, Tideways, AppSignal, Datadog.
- **Observability platforms** stack APM, logs, traces, errors, sessions, and frontend analytics into one dashboard. Think Datadog, Sentry (these days), PostHog.

## Commercial services

Often recommended in the PHP space.

### Tideways

Built for PHP, by PHP people, license via `php.ini`, sensible defaults. If you want one tool that profiles, monitors, and tracks errors without sprawling into ten product categories, this is the one we point at most often. No integration yet, but planned.

:ContentLink{href="/integrations/apm/tideways" text="Tideways on fortrabbit" prefix="docs"} · [tideways.com](https://tideways.com)

### Flare

Error tracking for Laravel, made by the people behind Spatie packages. Narrow focus, very polished. No integration yet, considering it.

:ContentLink{href="/integrations/apm/flare" text="Flare on fortrabbit" prefix="docs"} · [flareapp.io](https://flareapp.io)

### Sentry

Outgrew "error tracker" years ago. Now it does traces, profiling, and session replay too. The free tier is generous. Multi-stack tracking. No integration planned for now.

[sentry.io](https://sentry.io)

### Blackfire

The profiler from the Symfony world, nowadays integrated with Upsun.

:ContentLink{href="/integrations/apm/blackfire" text="Blackfire on fortrabbit" prefix="docs"} · [blackfire.io](https://blackfire.io)

### New Relic

The long-standing enterprise option. Capable and broad, a natural fit for a team that already runs on it. Expensive.

:ContentLink{href="/integrations/apm/new-relic" text="New Relic on fortrabbit" prefix="docs"} · [newrelic.com](https://newrelic.com)

### Datadog

An observability platform for large, multi-stack setups.

:ContentLink{href="/integrations/apm/datadog" text="Datadog on fortrabbit" prefix="docs"} · [datadoghq.com](https://www.datadoghq.com)

## Open source

### PHP SPX

A self-contained native profiler with a clean web UI. No agent, no SaaS, only a PHP extension and a browser tab.

[github.com/NoiseByNorthwest/php-spx](https://github.com/NoiseByNorthwest/php-spx)

### Excimer

A low-overhead sampling profiler from Wikimedia.

[github.com/wikimedia/mediawiki-php-excimer](https://github.com/wikimedia/mediawiki-php-excimer)

### SigNoz

An OTel-native, self-hostable backend for traces, metrics, and logs. The open-source, self-hosted angle on the same idea as Datadog.

:ContentLink{href="/integrations/apm/signoz" text="SigNoz on fortrabbit" prefix="docs"} · [signoz.io](https://signoz.io)

## Privacy

Every observability tool ships data off your server. The more tools you wire in, the more places URLs, query parameters, headers, and user IDs end up. Keep that in mind.

## Overhead

Modern APM agents should stay under a few percent of CPU. Stacking several of them — APM, error tracker, frontend analytics, session replay — adds up, so it is worth measuring before and after.

## OpenTelemetry

[OpenTelemetry](https://opentelemetry.io) is a vendor-neutral standard for traces, metrics, and logs. The promise of OTel is that you can swap your backend without re-instrumenting your code.

## Our take

We have always shown basic resource metrics — CPU, memory, request counts. The new dashboard will improve on this considerably (not ready yet), with much better discoverability of issues at platform level. See when an environment is unhealthy, what changed, and where to look first. This will be good for most common use cases, but not replace an APM tool.

We can not allow arbitrary code to be installed on the platform. Our architecture is designed to serve web applications, not to run as a multi-app server hosting background daemons next to your code. That shapes which integrations are technically possible at all.

We will consider carefully which commercial services to support. Each integration is a long-term commitment: we maintain the glue, document it. We would rather support a small set of integrations well than a long list of half-working ones.

---

- [APM integrations](/integrations/apm)
- [People problems](/dev/performance/people-problems)