SOC 2 Type 2 audited. Sandboxed execution. No model training on your code. Deploy in SaaS, your VPC, or fully air-gapped. Every commitment is contractual.
Fastest to start. Code is accessed via ephemeral, read-only clones in an isolated sandbox. Deleted at end of run.
We deploy the agent into your AWS / GCP / Azure account. Your IAM, your network, your logs. We provide the control plane.
For regulated environments. Standalone install, no egress required. Model weights delivered as signed artifacts.
| Data | What we do |
|---|---|
| Source code | Read in sandbox only. Not persisted beyond the run. Never used to train models. |
| Code sent to LLMs | Only the functions under analysis are passed to the model, not your full codebase. We use enterprise API agreements with zero training rights. Your code is not used to improve any model. |
| Runtime inputs (traces) | Stored encrypted. Scrubbed on request. Used only for benchmarking and regression tests. |
| Benchmark results | Stored in your tenant. Exportable. Retained per your contract. |
| Secrets / credentials | Never stored. Sandbox has no egress to secret stores. |
| PII / customer data | Not needed. Benchmarks run on synthetic or scrubbed inputs. |
Full subprocessor list available on request.
Responsible disclosure goes to security@codeflash.ai. We respond within one business day. We do not pursue legal action against researchers acting in good faith under our disclosure policy.
Pre-reviewed security packet available on request. SOC 2 Type 2 report available under NDA. We turn most security questionnaires around in 48 hours.