Continuous profiling without the platform to run it
Scheduled sampling, analysed automatically, so a regression shows up as a finding rather than as a complaint.
Continuous profiling normally means adopting a platform: a collector, a storage tier, retention policy, and somebody to own it. That is a reasonable trade at a certain size and a bad one below it. flamelens schedules the sampling and does the reading, and there is nothing for you to operate.
Scheduled, not constant
Sampling on a schedule catches drift without a permanent tax.
- Agents record on the interval you set
- Each recording is analysed like any upload
Live scans where they are safe
Attaching to a running JVM, with the boundary drawn explicitly.
- Targets validated before any connection is made
- Python and native targets are upload-only, by design
Findings over time
The point is the change, not the snapshot.
- Report history per service
- A new finding is the signal worth acting on
Nothing to operate
No collector, no storage tier, no retention policy of your own.
- The agent is one binary
- Reports live in your account
Frequently asked questions
How is this different from uploading a profile myself?
Only in who remembers to do it. Manual uploads answer a question you already had; scheduled sampling is what surfaces the question. The analysis is identical.
Can every runtime be scanned live?
No. Live scanning works by attaching to a management interface the JVM exposes. Python and native targets have no equivalent that can be reached safely, so they stay upload-only — a limit worth stating plainly rather than discovering later.
What stops the agent being pointed at something it should not reach?
Every target is validated before a connection is attempted. An endpoint that accepts an arbitrary host to connect to is a server-side request forgery waiting to happen, so the validation is a boundary rather than a convenience check.
Upload a profile. Leave with a plan.
Free to start — 10MB recordings, full reports, no card.
Create a free account