Integrations
Start with your logs. Add context as you need it.
Keep the data on your own infrastructure, with one Slowpoke server and an agent on each monitored machine.
What it reads
It reads what your servers already write
You change no code to start, and nothing sits between your users and your application. Any backend language works through access logs and database statistics; framework packages and OpenTelemetry add the code origins on top.
nginx 路 Apache 路 Caddy
Requests per endpoint, slow requests, errors and a latency histogram, all from the access log you already have.
MySQL 路 MariaDB
Slow log and performance_schema, with samples and real EXPLAIN plans. MySQL 5.7 and 8.
PostgreSQL
pg_stat_statements with EXPLAIN; generic plans for statements stored without values, on PostgreSQL 16 and newer.
Redis
Hit rate, memory, evictions and the slow command log as timers, with the two lines to turn that log on (and the two to turn it off).
php-fpm slow log
Exactly where PHP workers stall, down to the file and the line, without a profiler in production. One box per pool when a machine runs several.
The machine (/proc)
CPU, iowait, steal, memory, swap, OOM kills, disk, network, top processes, and the process behind each connection.
Jobs and schedules
Queued jobs and scheduled commands, with their queue, their run count and what each run costs.
OpenTelemetry (OTLP)
Routes, queries per route and N+1 from any backend language, received directly by the agent.
Framework packages
The file and line of every query. Packages for Laravel, Symfony, Python and Node.js. Check the documentation for supported framework versions and installation instructions.
StatsD 路 Prometheus
Your own counters, gauges and timers, next to everything else on the same scale.
RabbitMQ
Queue depth, consumers and rates from the management API, with the seconds a message waits before anyone picks it up.
Elasticsearch 路 OpenSearch
Seconds spent searching, thread pool rejections, the heap of the worst node, and cluster status as a number.
MongoDB
The mongod JSON log: slow operations, collection scans, documents examined, timed per collection and operation. The filter never leaves the machine.
Deploys
A webhook or a CLI call marks a deploy; every one gets a verdict against the hour before it.
Everything else
Use StatsD, Prometheus or OpenTelemetry for additional application measurements. Check the documentation for the supported signals and configuration.
Know what is covered and what is missing.
Each source reports its status. Missing request timings, unavailable slow logs or database permissions appear as explicit gaps, with configuration guidance. Optional server changes require your acceptance; review the suggested configuration before applying it.
Bring the evidence to your coding assistant.
Missions are available over MCP, including query plans and code origins where collected. Your assistant can use that context to propose a fix. Slowpoke runs no model internally; the assistant you choose controls what is sent to its provider. You review and ship the changes.
Keep work in the tools your team uses.
Turn missions into issues in GitHub, GitLab, Linear, Jira, Asana, Trello or ClickUp. Send digests and deploy verdicts through Slack, Discord, Teams, Telegram or webhooks.
Designed to keep monitoring overhead small.
The agent reads existing logs and statistics, with bounded buffers and a soft memory target. It reports its own resource usage so you can measure the impact on your machines. Query values are redacted by default, and query plans are obtained locally from statements already observed.
Deploy on your own infrastructure.
Install the panel on a Docker host, create a project and use its generated command to connect an agent. Follow the coverage checks to enable the sources you need. Add an optional framework package for query origins. Read installation and integration docs.