Gardera runs its autonomous security engineer on Opper
- Industry
- Security
- Data policy
- Zero data retention
“Our customers' code is never stored. With Opper, that holds all the way to the model layer, at billions of tokens a day.”
About Gardera
A security engineer that runs your AppSec on autopilot.
Gardera is a Stockholm security company founded in 2025 by a team of AppSec practitioners. Its autonomous security engineer reads a codebase the way an engineer would, proves which findings are genuinely exploitable, and opens the pull requests that fix them. Gardera also runs AI penetration testing: swarms of offensive agents that attack customers' applications the way a real adversary would, chaining weaknesses together to show what an attacker could actually reach. It plugs into the tools developers already use, from GitHub and GitLab to CI pipelines and AI coding assistants, with a two-minute setup.
The Challenge
Reasoning over whole codebases, without storing a line of them.
Every Gardera finding is the product of reasoning over real code: following data flow across files, confirming a vulnerability is reachable, drafting the fix. Its AI pentests add a swarm of attacking agents working in parallel, each reasoning through its own chain of model calls. That means long-context model calls, at a volume that grows with every repository and application a customer connects.
Two constraints came with it. Gardera promises customers their code is never stored, so the model layer had to meet the same standard. And the team's time belongs in the product, not in provider integrations, quotas and failover.
The Solution
One EU-hosted AI gateway behind every finding.
Gardera runs its inference on Opper, from context and reachability analysis and its AI pentesting agents to the generated fix. Every call goes to frontier models through a single EU-hosted AI gateway, with no model infrastructure of Gardera's own to run.
- EU-hosted, with the gateway in AWS Stockholm and EU model routes, under one DPA that covers every model.
- Zero data retention on the routes that handle customer code: neither Opper nor the model provider stores prompts or completions.
- Model freedom from the control plane, so any step can move to a better model when one ships, with no integration rewrite.
- Automatic fallback across providers, so a degraded endpoint fails over instead of stalling a scan.
Results
Frontier models at production scale, and not a line of customer code stored.
Gardera went from first calls to production in days, with no integration project. The whole pipeline, from context and reachability through AI pentesting to the fix, runs through Opper with the data promise intact.
- Engineering time goes into the product, not into model infrastructure.
- Headroom as customers connect more repositories, with routing, fallback and data residency handled at the gateway.