Check 1
Infrastructure reconnaissance
Confirm the API is reachable and collect DNS, TLS, and entry-response clues before later checks.
Check whether your API relay swaps, injects, or leaks data — all in one audit
Paste an OpenAI-compatible endpoint, API key, and model ID. LMSpeed runs all 13 security probes in one pass and returns a redacted shareable report.
- All-in-one
- 13 checks
- Risk surfaces
- 9
- Output
- Redacted report
Latest 100/100 API Security Audits
The newest relay audit reports where endpoint profile, model identity, prompt safety, and response integrity all scored 100.
- sub.callai.onegpt-5.6-solReport timeAug 18Report timeAug 18848486100
- sub.callai.onegpt-5.6-solReport timeAug 18Report timeAug 18707280100
- ai.databyte.co.iddatabyte-m1Report timeAug 18Report timeAug 187284100100
- ai.databyte.co.iddeepseek-v4-flashReport timeAug 18Report timeAug 1810084100100
- ai.databyte.co.idMiniMax-M3Report timeAug 18Report timeAug 18768478100
- tokengate-cqt9ivzs.manus.spaceclaude-opus-5Report timeAug 17Report timeAug 17648465100
- router.bynara.idqwen-3.8-max-freeReport timeAug 17Report timeAug 17846686100
- 87.120.187.161claude-fable-5Report timeAug 17Report timeAug 1772727273
- muyuan.doGLM-5.2Report timeAug 17Report timeAug 17768468100
- code28.ccwu.ccGPT-5.6 TerraReport timeAug 17Report timeAug 17848486100
More than a speed test: inspect whether the relay path was tampered with
lmspeed puts model identity, prompt leakage, context boundaries, error leakage, and stream integrity into one security comparison table, so you can baseline a relay before wiring it into production.
| Dimension | lmspeed | hvoy.ai | cctest.ai |
|---|---|---|---|
| Token injectionCompare actual token usage with the expected count | Covered | Not covered | Covered |
| Prompt extractionProbe hidden system prompt leakage | Covered | Not covered | Not covered |
| Identity substitutionDetect whether Claude is actually answered by another model | Covered | Covered | Not covered |
| Jailbreak defenseCheck common jailbreak vectors | Covered | Not covered | Not covered |
| Context truncationFind the real context-window boundary | Covered | Not covered | Not covered |
| Tool-call rewrite (AC-1.a)Detect rewritten package commands and tool arguments | Covered | Not covered | Not covered |
| Error response leakage (AC-2)Probe credentials, paths, and internal field leakage | Covered | Not covered | Not covered |
| Stream integrity (SSE)Validate event types, usage, and thinking signatures | Covered | Covered | Not covered |
| Web3 injectionCheck whether signing context is polluted by the relay layer | Covered | Not covered | Not covered |
| Channel fingerprintProtobuf signatures and multimodal interpretation checks | In designSoon | Not covered | Covered |
CoveredCoveredNot coveredNot coveredIn designSoonIn design
How the 13-check audit breaks down relay risk
Each check keeps public evidence redacted: you can see where the path looks suspicious without publishing API keys, system prompts, or internal paths.
Threat categories are based on Liu et al., "Your Agent Is Mine" (arXiv:2604.08407)
Check 2
Model list
Read the public model catalog and check whether the requested model is actually listed.
Check 3
Token injection
Compare billed or reported input tokens with the expected count to find a hidden system prompt.
Check 4
Prompt extraction
Try verbatim, translation, and JSON-continuation probes to extract hidden system instructions.
Check 5
Instruction conflict and model identity
Check whether a high-priority instruction is overridden and whether the answering identity matches the request.
Check 6
Jailbreak and identity leakage
Use diagnostic, encoding, and role-play probes to see whether identity or policy leaks.
Check 7
Context window
Increase context until the usable boundary appears, not only the advertised window.
Check 8
Tool-call rewrite
Detect whether package-install commands are rewritten on the return path.
Check 9
Error response leakage
Trigger errors and look for credentials, upstream URLs, or internal paths.
Check 10
Stream integrity
Validate SSE event structure and whether the streamed model name matches the request.
Check 11
Web3 prompt injection
Check whether wallet-safety probes produce private-key or signing advice.
Check 12
NewAPI and LLM gateway fingerprint
Identify NewAPI and other LLM gateway stacks behind the relay so later anomalies are easier to explain.
Check 13
Latency variance
Repeat the same request and look for queues, extra hops, or silent model switching.
FAQ
What is the all-in-one API relay audit?
It is one run of 13 checks on an OpenAI-compatible API relay. LMSpeed looks for model swaps, hidden prompts, context truncation, tool-call rewrites, error leaks, and stream problems, then returns a redacted report.
Why run 13 checks instead of a speed test?
A speed test measures wait time and throughput. These 13 checks look at whether the relay path was tampered with. You can still open any single check if you only need one risk.
Do public reports include API keys or system prompts?
No. Public reports stay redacted. You can see that a leak or swap happened without LMSpeed publishing secrets.
Can I run one check instead of all 13?
Yes. Ten checks have their own pages. Reachability, model list, and gateway fingerprint stay inside the full 13-check suite.