Overview
PentAGI uses specialized AI agents for autonomous penetration testing. This API provides access to agent logs, assistants, and various execution logs including terminal output, web searches, and inter-agent communications.
Assistants
Assistants are interactive AI agents that can answer questions and perform follow-up tasks within a flow.
Assistant Type
type Assistant {
id: ID!
title: String!
status: StatusType!
provider: Provider!
flowId: ID!
useAgents: Boolean!
createdAt: Time!
updatedAt: Time!
}
Whether the assistant can delegate to specialized agents or works independently
Query Assistants
query GetAssistants($flowId: ID!) {
assistants(flowId: $flowId) {
id
title
status
provider {
name
type
}
useAgents
createdAt
}
}
Create Assistant
mutation CreateAssistant(
$flowId: ID!
$provider: String!
$input: String!
$useAgents: Boolean!
) {
createAssistant(
flowId: $flowId
modelProvider: $provider
input: $input
useAgents: $useAgents
) {
flow {
id
updatedAt
}
assistant {
id
title
status
}
}
}
{
"flowId": "1",
"provider": "default-openai",
"input": "Explain the SQL injection vulnerability found in the login form",
"useAgents": false
}
Call Assistant
Send a follow-up message to an existing assistant.
mutation CallAssistant(
$flowId: ID!
$assistantId: ID!
$input: String!
$useAgents: Boolean!
) {
callAssistant(
flowId: $flowId
assistantId: $assistantId
input: $input
useAgents: $useAgents
)
}
Stop Assistant
mutation StopAssistant($flowId: ID!, $assistantId: ID!) {
stopAssistant(flowId: $flowId, assistantId: $assistantId) {
id
status
}
}
Delete Assistant
mutation DeleteAssistant($flowId: ID!, $assistantId: ID!) {
deleteAssistant(flowId: $flowId, assistantId: $assistantId)
}
Agent Logs
Agent logs track inter-agent communications and task delegations.
AgentLog Type
type AgentLog {
id: ID!
initiator: AgentType!
executor: AgentType!
task: String!
result: String!
flowId: ID!
taskId: ID
subtaskId: ID
createdAt: Time!
}
Agent that initiated the task delegation
Agent that executed the task
Query Agent Logs
query GetAgentLogs($flowId: ID!) {
agentLogs(flowId: $flowId) {
id
initiator
executor
task
result
taskId
subtaskId
createdAt
}
}
{
"data": {
"agentLogs": [
{
"id": "log-001",
"initiator": "primary_agent",
"executor": "searcher",
"task": "Search for recent CVEs affecting Apache 2.4.41",
"result": "Found 3 relevant CVEs: CVE-2021-44790 (critical), CVE-2021-44224 (high), CVE-2021-26691 (medium)",
"taskId": "task-001",
"subtaskId": "subtask-002",
"createdAt": "2024-02-20T10:33:15Z"
}
]
}
}
Subscribe to Agent Logs
subscription AgentLogAdded($flowId: ID!) {
agentLogAdded(flowId: $flowId) {
id
initiator
executor
task
result
createdAt
}
}
Terminal Logs
Terminal logs capture all command executions and their output.
TerminalLog Type
type TerminalLog {
id: ID!
flowId: ID!
taskId: ID
subtaskId: ID
type: TerminalLogType!
text: String!
terminal: ID!
createdAt: Time!
}
Stream type: stdin, stdout, or stderr
ID of the terminal (container) where the command was executed
Query Terminal Logs
query GetTerminalLogs($flowId: ID!) {
terminalLogs(flowId: $flowId) {
id
type
text
terminal
taskId
subtaskId
createdAt
}
}
{
"data": {
"terminalLogs": [
{
"id": "tlog-001",
"type": "stdin",
"text": "nmap -sV -p- target.example.com",
"terminal": "term-001",
"taskId": "task-001",
"subtaskId": "subtask-001",
"createdAt": "2024-02-20T10:31:00Z"
},
{
"id": "tlog-002",
"type": "stdout",
"text": "Starting Nmap 7.94 ( https://nmap.org )\nNmap scan report for target.example.com (192.168.1.100)\nHost is up (0.0012s latency).",
"terminal": "term-001",
"taskId": "task-001",
"subtaskId": "subtask-001",
"createdAt": "2024-02-20T10:31:01Z"
}
]
}
}
Subscribe to Terminal Logs
subscription TerminalLogAdded($flowId: ID!) {
terminalLogAdded(flowId: $flowId) {
id
type
text
terminal
createdAt
}
}
Message Logs
Message logs track agent communications, thoughts, and results.
MessageLog Type
type MessageLog {
id: ID!
type: MessageLogType!
message: String!
thinking: String
result: String!
resultFormat: ResultFormat!
flowId: ID!
taskId: ID
subtaskId: ID
createdAt: Time!
}
Message type: answer, report, thoughts, browser, terminal, file, search, advice, ask, input, done
AI agent’s reasoning process (if available)
Format of the result: plain, markdown, or terminal
Query Message Logs
query GetMessageLogs($flowId: ID!) {
messageLogs(flowId: $flowId) {
id
type
message
thinking
result
resultFormat
taskId
subtaskId
createdAt
}
}
Search Logs
Search logs track web searches performed by agents.
SearchLog Type
type SearchLog {
id: ID!
initiator: AgentType!
executor: AgentType!
engine: String!
query: String!
result: String!
flowId: ID!
taskId: ID
subtaskId: ID
createdAt: Time!
}
Search engine used: “tavily”, “duckduckgo”, “google”, “perplexity”, etc.
Query Search Logs
query GetSearchLogs($flowId: ID!) {
searchLogs(flowId: $flowId) {
id
initiator
executor
engine
query
result
createdAt
}
}
Vector Store Logs
Vector store logs track memory operations.
VectorStoreLog Type
type VectorStoreLog {
id: ID!
initiator: AgentType!
executor: AgentType!
filter: String!
query: String!
action: VectorStoreAction!
result: String!
flowId: ID!
taskId: ID
subtaskId: ID
createdAt: Time!
}
action
VectorStoreAction
required
Action performed: retrieve or store
Screenshots
Screenshots captured during browser-based testing.
Screenshot Type
type Screenshot {
id: ID!
flowId: ID!
taskId: ID
subtaskId: ID
name: String!
url: String!
createdAt: Time!
}
Query Screenshots
query GetScreenshots($flowId: ID!) {
screenshots(flowId: $flowId) {
id
name
url
taskId
subtaskId
createdAt
}
}
Agent Types
Available AI agent types:
primary_agent - Main orchestration agentassistant - Interactive question-answering agentpentester - Security testing specialistcoder - Script and exploit developmentinstaller - Tool installation and setupsearcher - Web search and researchmemorist - Memory storage and retrievaladviser - Strategic advice and planninggenerator - Task decompositionrefiner - Result refinementreflector - Analysis and insightsenricher - Context enrichmentreporter - Report generationsummarizer - Summary creationtool_call_fixer - Tool call error recovery