An API for authorized security work, run by Uplink Switch LLC.
PentestLLM sells API access to language models for security testing: 11 builds set up for security work and 31 abliterated builds on 2026-11-11, on one prepaid key. Every listed model speaks the OpenAI wire format, so the official SDKs and any coding agent that takes a base URL connect without a plugin.
curl https://api.pentestllm.ai/v1/chat/completions \
-H "Authorization: Bearer sk-ptl-live-…" \
-H "Content-Type: application/json" \
-d '{"model": "cyber-frontier", "messages": [{"role": "user", "content": "Hello"}]}'Who it is for
Penetration testers, red teams, bug bounty hunters, CTF players and the engineers who defend against them. The work is triaging scanner output, explaining a CVE, reviewing code for vulnerabilities, drafting report findings, writing detection rules and solving CTF challenges. Stock assistants often decline these questions; the builds here are chosen because they decline fewer of them. Refusals are reduced, not removed.
The service is for authorized testing only: systems you own, or systems you have written permission to test. The acceptable use policy says what is not allowed, and it is enforced. You must be 18 or older.
Who runs it
Uplink Switch LLC is a Wyoming limited liability company. It runs the gateway, the console where keys and top-ups live, and the API at api.pentestllm.ai. It does not train models. Each request goes to a host that serves the build, the answer comes back, and the gateway records what was billed. The hosts are not named on this site. The models are, each with its price and its result from the daily check.
What the daily check does
Once a day every model on sale gets the same fixed test prompt and is asked for a single output token. The input token count comes from the model’s own tokenizer, so a genuine build reports the same count day after day. A swapped model, or instructions slipped into the prompt on the way, shows up as a different count.
On 2026-11-11, 36 of 39 listed models held their count. The daily check page lists every model with its verdict: steady since a date, changed on a date, or not checked yet. It is evidence, not proof. Two builds that share a tokenizer report the same count, so the check catches a different model or an added prompt, not a swap between two tunes of one base.
36of 39 steady
One cell per listed model; hover a cell for its verdict.
How billing works
You top up a prepaid balance in the console from $10. Each request settles against it at the input and output prices on its model’s row of the pricing page, counted to the token. Larger top-ups add bonus credit: 5% from $200, 10% from $500, 15% from $1,000.
A request that fails before the model answers costs nothing. When the balance can no longer cover a request, the API answers 402 and stops. No subscription is required and credit has no expiry date. The refund policy covers billing errors.
- Input, 8,000 tokens at $0.40 per million
- $0.0032
- Output, 1,000 tokens at $2.40 per million
- $0.0024
- One turn
- $0.0056
- 1,000 turns
- $5.60
What we keep
One usage record per request: the request id, the model, token counts and the charge. Next to it, SHA-256 fingerprints of the request and the response, never the text. Prompts and completions pass through and are not written anywhere, unless you turn on replay for a key, which keeps that key’s request bodies for 7 days. The hosts that run the models receive the request and the response for the calls they serve. The privacy page has the full list.
Reporting a problem
Abuse of the service: abuse@pentestllm.ai, no account needed. A vulnerability in the service itself: the disclosure policy. Everything else: hello@pentestllm.ai or the contact page.
Company
Uplink Switch LLC, 1309 Coffeen Ave STE 20866, Sheridan, WY 82801. Keys and top-ups live in the console.