What a fixed-price pentest of one web app or one AI agent includes
Teams we work with in Da Nang ship a web app or an agent, then a customer's security questionnaire asks for a recent pentest. The quotes that come back are often five figures and a calendar measured in weeks. A newer set of sites will sell a scan the same day for about $99 to $495. This page is a straight read of one service that sits between those two, Observ, because it publishes the price, the paperwork, and the work it will turn down. Prices below are what that site showed on 26 October 2026.
What are you actually buying?
One authorized test of one thing you name, plus the paperwork a customer's security team already knows how to file. There are two products. A second target is a second test, at the same price.
| Test | What it covers |
|---|---|
| Host pentest, $3,995 | One hostname and port, such as a web app or an API. Testing follows OWASP guidance and stays inside the signed scope. |
| AI agent pentest, $5,995 | One AI agent or LLM app, exercised through the interface your users already use, signed in as a test user you provide. The published categories are prompt injection, tool and permission misuse, data the agent should not hand out, unsafe actions, output handling, and runaway cost. The model provider is out of scope. Detail is on Observ's agent pentest page. |
Both include a report, a signed attestation letter, a debrief with your engineers, and one retest of the findings you have fixed, if you ask inside 90 days. The retest checks those fixes. It is not a second full test.
How does the engagement actually run?
The form on the site asks for a conversation. It does not reserve a window, and it does not start a test. A deposit reserves the window. Written authorization starts the test. Observ's site does not take payment. The deposit amount, and when it is refunded, sit in the engagement they send, which is also where the terms point you.
| Step | What happens |
|---|---|
| Scope | You agree the host or the agent, the dates, and the rules of engagement, and both sides sign an NDA. For an agent, you also provide a test user. |
| Test | Starts within 10 business days of the signed scope. Nothing outside that scope is in play. |
| Report | Within 5 business days of testing. A person on their team has confirmed every finding before it ships. |
| Debrief | A call with your engineers, finding by finding, including the fix. |
| Retest | Inside 90 days, each fixed finding is marked closed or still open. |
What is in the report, and what is the letter?
The report is for two readers. Leadership gets an executive summary. Engineers get findings ranked by CVSS, each with where it lives, evidence the team can reproduce, and the change to make. On an agent test, that evidence is the transcript and the tool calls. The example findings on Observ's own site are illustrations of that shape. They are not results from your host.
The attestation letter is the page you can forward. It states what was tested, when, and the results, so an auditor or a customer's security review can see that an independent test happened. Observ says plainly that it is not a SOC 2 or ISO 27001 auditor, and that the letter is not a certification. Those audits stay separate work, done by someone else.
What will they refuse?
The gate is written authorization. If they cannot see it, they do not start. The published refusal list is short, and it is the part worth reading before you book:
- No written authorization for the host or the agent.
- A third party you do not control. On an agent test, that includes the model provider.
- A request for exploit code, a weaponized payload, or prompts to reuse somewhere else.
- A production host, or a production agent that moves money or sends messages, with no test window. The usual path is a staging environment. The other is a signed acceptance of production risk.
What they need from you is the hostname and port, or the agent and the tools it can call, an authorization contact and a signature, a window, and either staging or that signed acceptance.
How does this compare with a same-day scan?
On 26 October 2026 the cheap end of this market was easy to find, and it is a real product. TurboPentest lists a pentest from $99 a target, with a report in hours and a signed attestation letter. Its attestation page says the letter is not a SOC 2 or ISO 27001 certificate. Its pricing page says the $99 tier is enough when an auditor only asks for an annual penetration test. AssurePort lists one web pentest from $399, with a PDF and JSON report. FastPentests lists a $495 test aimed at a security questionnaire, and says a manual pass checks the automated findings. FastPentests also describes the traditional quote it is pricing against as $5,000 to $25,000.
Read who confirms the finding. TurboPentest says its AI agents verify findings. Observ says a person on the team confirms every finding before the report reaches you, and the test does not start until the scope is signed. That is a slower, more expensive product. It is also the one that will decline the work when the authorization is missing.
Who should buy which option?
| Option | Fit |
|---|---|
| A same-day scan, about $99 to $495 |
Best for: a questionnaire that landed today, on one web app, where you can accept a productized scan and you will read the letter before you attach it. Honest limit: depth, human review, and what the letter is allowed to stand in for all vary by vendor. A low price does not tell you which of those you bought. |
| A traditional pentest firm |
Best for: a multi-application annual program, a regulated scope, or a customer who has named a methodology and a firm class. Honest limit: the public band sellers quote against is five figures, and the calendar is often weeks. Overkill when you have one host and a questionnaire. |
| Observ |
Best for: one web app, one API, or one AI agent, where you want a signed scope, a person confirming findings, a letter a customer's security team will recognise, and a retest of the fixes, at a price published on the site. Honest limit: one target per engagement, ten business days to start, and the letter is not a certification. The marketing site does not take payment. You are also buying from Cipher Projects, whose founder directs Tech Ops Asia. Treat this page as a description from a related company, and judge the engagement on its own terms. |
| Tech Ops Asia |
Best for: a company in Vietnam that needs the surrounding work: how personal-data rules apply to the AI tools already in use, a workshop on a real workflow, or engineers in Da Nang. Honest limit: we do not sell, scope, or run pentests, and we are not a law firm. If the only question is the test, the link above is the answer. |
What does a pentest leave untouched in Vietnam?
A pentest answers a security questionnaire. It does not file a dossier, and it does not decide whether staff or customer data may go into a model. Those are operating-controls questions under Vietnam's personal-data rules, with administrative fines attached since Decree 330. Our read of that layer is in how Vietnam law treats company AI tools and in what Decree 330 changed for a normal company. Counsel still files. A letter from a pentest firm does not replace either of those.
Booking the test, or the Vietnam side of the same problem?
The test is booked with Observ. If the harder part is how your company in Vietnam handles the data those systems touch, talk to us.
FAQ
Is Observ a Tech Ops Asia service?
Observ is the pentesting service of Cipher Projects. Keith Vaughan directs Tech Ops Asia and founded Cipher Projects. This page describes the test. Tech Ops Asia does not scope it, run it, or invoice it.
What does $3,995 include?
One authorized pentest of one hostname and port, a report, a signed attestation letter, a debrief call, and one retest of the findings you have fixed if you ask inside 90 days of the report. The retest checks those fixes. It is not a second full test. A second host is a second test at the same price. Figures are as Observ published them on 26 October 2026.
Is the attestation letter a SOC 2 or ISO certificate?
The letter states what was tested, when, and the results. You can hand it to an auditor or a customer's security review. Observ is not a SOC 2 or ISO 27001 auditor, and the letter is not a certification.
Will they test a production host?
Yes, if you schedule a window or sign an acceptance of production risk. They turn down a production host with no window. They do not start without written authorization, and they do not test a third party you do not control.
Do they deliver exploit code?
They turn that down, including a request for prompts to reuse on some other agent. Each finding in the report has evidence your team can reproduce, and the fix.
What is the difference between the host test and the agent test?
The host test is one hostname and port, at $3,995, following OWASP testing guidance. The agent test is one AI agent or LLM application, at $5,995, following the OWASP Top 10 for LLM Applications, signed in as a test user you provide. The model provider is out of scope. The report, the letter, the debrief, and the retest are the same shape.