Autonomous pentest console

An AI crew runs the pentest. You hold the switch.

Hackeris.lv is a console for authorized security testing. You name the targets and the rules. A crew of AI agents plans the work, runs the commands on an isolated Kali machine, proves each finding and writes it down.

For targets that you own or may test in writing. It runs on your own server, with your own AI model keys.

Look around a static demo — fictional data, no sign-in, nothing saved.

# Test: Example shop, run 3. Kali machine kali-01, VPN lab-eu

nmap -sV -Pn shop.example.test

22/tcp open ssh OpenSSH 9.6

443/tcp open https nginx 1.26

note saved: two services, the web shop is the way in

ffuf -u https://shop.example.test/FUZZ -w common.txt -rate 2

/admin [Status: 302]

/api/v1 [Status: 200]

curl -s https://payments.example.test/health

blocked: payments.example.test is a blocked target

TestsExample shopRun 3kali-01VPN lab-euRunning
Crew
5 agents2 running
$0.72
PlannerSonnet 5.5Running

Plan the test of shop.example.test and split the work between the workers.

Recon workerHaiku 5.5Done

Map the open ports and the services of the two targets in scope.

Web workerSonnet 5.5Running

Test the sign-in and the order API for access control flaws.

VerifierSonnet 5.5Needs you

Run the proof of the order API finding again, from a clean session.

ScribeHaiku 5.5Waiting

Keep the notes and write the run report.

skills in the library
54
Web, API, Active Directory, cloud, privilege escalation.
roles in a crew
5
Planner, worker, verifier, triage and scribe.
bug bounty platforms read
3
HackerOne, Intigriti and YesWeHack.
reports sent by the app
0
You copy every report yourself. No setting changes this.

01Crew

A crew with roles, not one long chat

A planner splits the Test. Workers run the commands. A verifier runs the proof of each finding again, and a scribe keeps the notes. Each agent shows its state, its cost and its time while it works.

  • State as a colour, a word and an icon
  • Estimated cost and time on every card
  • One queue for everything that waits for you

02Approvals

A command waits for a rule or for you

Choose “Ask me before each command” and the run stops at every step until a person says yes. A Test that is bound to a bug bounty program also runs behind a wall: scope, request rate and blocked targets are checked for each request.

  • Blocked targets stop a run before it starts
  • One switch per Kali machine: commands allowed, or not
  • AI stop ends all AI work in the workspace at once
ApprovalVerifier, run 3
Needs you

The verifier wants to run one command on kali-01.

curl -s -H "X-Bug-Bounty: your-handle" https://shop.example.test/api/v1/orders/1042
  • Blocked targetsnone touched
Waits for you
Scope of the programExample Retail, read 2 h ago
TargetKindRule
shop.example.testDomainIn scope
*.api.example.testWildcardIn scope
payments.example.testDomainBlocked
192.0.2.0/28NetworkOut of scope

03Findings

A finding arrives with its proof

A finding carries its evidence in order: the request and the response, a screenshot, a result file. The verifier confirms it or leaves it pending with a note. The Report Workshop helps you write and check the report, and you copy the text yourself.

  • Severity with a CVSS score, status as a word
  • The verifier never dismisses a finding
  • Five ticks and your typed name before a report leaves
FindingsExample shop, 3 of 3
  • High8.1The order API returns the orders of other customersshop.example.test/api/v1/ordersConfirmed
  • Medium5.3A password reset link works more than onceshop.example.test/resetPending
  • Low3.1The API shows the server version in an error pageshop.example.test/api/v1Duplicate
The order API finding, step by step
  • Found by the web workerRun 314:02
  • Evidence attachedrequest and response2 items14:03
  • The verifier ran the proof againconfirmedVerifier14:09
  • Report draft in the Report WorkshopReady score 8214:20
  • You check the text and copy itYou

04AI use

The cost is on screen, not in a bill later

You add an AI model with its key and its prices. The console counts tokens and works out the estimated cost of every run. A daily AI budget warns or stops the AI at the limit.

  • Tokens and estimated cost for each run and each day
  • A daily AI budget for one UTC day
  • Your models, your keys, your prices

AI use

Tokens and runs for each day, 30 days · 1 Sep – 30 Sep

AI use
Tokens
2,729,876
Runs
55
Estimated cost
$10.92
Sample tokens and runs for each day of 30 days
Tokens and runs for each day, 30 daysTokensRuns
1 Sep38,0001
2 Sep183,6764
3 Sep161,5143
4 Sep77,5142
5 Sep46,6761
6 Sep00
7 Sep00
8 Sep94,8662
9 Sep55,3521
10 Sep161,0844
11 Sep144,5703
12 Sep112,2182
13 Sep00
14 Sep00
15 Sep170,5983
16 Sep83,5702
17 Sep49,7041
18 Sep230,4924
19 Sep127,6263
20 Sep00
21 Sep00
22 Sep173,1964
23 Sep153,6543
24 Sep118,2742
25 Sep44,0561
26 Sep207,9004
27 Sep00
28 Sep00
29 Sep52,7321
30 Sep242,6044

What is in the console

Every part below is a page of the app. Nothing here is a plan.

  • Kali machinesThe commands of a Test run in an isolated Linux container, never on the server itself.
  • VPN profilesA machine reaches a lab network through a saved OpenVPN profile, with an isolation check first.
  • Skill librarySkills give the AI a method as plain text. 54 come with the app, and you can write your own.
  • WorkflowsAn order of steps that a Test can follow, so a method repeats the same way each time.
  • Shells and session replayOpen a terminal in any Kali machine. The saved output plays again as it happened.
  • Bug bounty programsThe app keeps a copy of a program: its policy, its scope and its rules. It only reads the platform.
  • SchedulesStart a Test by the clock, with a time zone, a catch-up rule and quiet hours.
  • AlertsA message to Discord or ntfy when a run needs you or when something is found.
  • Daily AI budgetA limit on the estimated cost of one day. At the limit the AI warns or stops.

From an empty workspace to a checked finding

  1. Add an AI modelPaste the key of a model and its prices. Pick the default model.
  2. Create a TestA target, a goal and the rules. Bind it to a bug bounty program when there is one.
  3. Start a runThe crew works on a Kali machine. Approve the commands, or let the rules decide.
  4. Check the findingsRead the evidence, reproduce the finding by hand, then copy the report.

What the app never does

A security tool earns trust by its limits. These are fixed in the product; no setting turns them off.

  • It never submits a report. You copy the text and paste it into the platform yourself.
  • It never writes to a bug bounty platform. It sends read requests only.
  • It never touches a blocked target. A run that would reach one stops before it starts.
  • It does not replace your judgment. Reproduce a finding by hand before you report it.

Test only what you may test

Read the rules of use first. Then sign in and start with one target, one goal and “Ask me before each command”.