POST
/api/v1/security/auto-attackStart an automated adversarial attack session
REGISTERS an automated adversarial attack session (TAP or IRT strategy) against a target model in a project. Creates an `auto_attack_sessions` row with `status: "pending"` and returns a sessionId — it does NOT execute the session: there is no worker that advances a stored session, so polling GET will keep returning `pending`. To run an adaptive red-team session today use the synchronous `POST /api/v1/security/adaptive` (BYOK target key). Requires `projectId` and `targetModel`; `goal` is required unless `useLibrary` is true. Optional `maxRounds` (1-20, default 3), `candidatesPerRound` (1-20, default 5), `strategy` (default `tap`), `categories`, `targetSystemPrompt`, `stopOnFirstSuccess`, `libraryCategory`, `libraryCount` (default 20). Returns 201; a failed insert is a hard 500 SESSION_PERSIST_FAILED rather than an unpollable id.
Authentication
Send Authorization: Bearer YOUR_API_KEY on every request. Generate API keys at /dashboard/settings/api-keys.
Request body required
Example
{
"projectId": "00000000-0000-0000-0000-000000000000",
"targetModel": "string",
"goal": "<Attack goal; required unless useLibrary=>",
"maxRounds": 1,
"candidatesPerRound": 1,
"categories": [
"string"
],
"strategy": "tap",
"targetSystemPrompt": "string",
"stopOnFirstSuccess": false,
"useLibrary": false,
"libraryCategory": "string",
"libraryCount": 1
}Schema
{
"application/json": {
"schema": {
"type": "object",
"properties": {
"projectId": {
"type": "string",
"format": "uuid",
"pattern": "^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$"
},
"targetModel": {
"type": "string",
"minLength": 1,
"maxLength": 200
},
"goal": {
"type": "string",
"maxLength": 20000,
"description": "Attack goal; required unless useLibrary=true."
},
"maxRounds": {
"type": "integer",
"minimum": 1,
"maximum": 20
},
"candidatesPerRound": {
"type": "integer",
"minimum": 1,
"maximum": 20
},
"categories": {
"maxItems": 50,
"type": "array",
"items": {
"type": "string",
"maxLength": 120
}
},
"strategy": {
"type": "string",
"enum": [
"tap",
"irt"
]
},
"targetSystemPrompt": {
"type": "string",
"maxLength": 100000
},
"stopOnFirstSuccess": {
"type": "boolean"
},
"useLibrary": {
"type": "boolean",
"description": "Test against the built-in jailbreak library instead of LLM-generated attacks."
},
"libraryCategory": {
"type": "string",
"maxLength": 120
},
"libraryCount": {
"type": "integer",
"minimum": 1,
"maximum": 500
}
},
"required": [
"projectId",
"targetModel"
],
"additionalProperties": false
}
}
}Response
201 example
{
"success": true
}All status codes
201Session created (sessionId + config).
400(no description)
401(no description)
403Forbidden — the credential is valid but lacks the API-key scope, member role, or plan entitlement this operation requires.
429(no description)
500Internal Server Error — SESSION_PERSIST_FAILED.
Code samples
cURL
curl -X POST \
https://evalguard.ai/api/v1/security/auto-attack \
-H "Authorization: Bearer $EVALGUARD_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "projectId": "00000000-0000-0000-0000-000000000000", "targetModel": "string", "goal": "<Attack goal; required unless useLibrary=>", "maxRounds": 1, "candidatesPerRound": 1, "categories": [ "string" ], "strategy": "tap", "targetSystemPrompt": "string", "stopOnFirstSuccess": false, "useLibrary": false, "libraryCategory": "string", "libraryCount": 1 }'TypeScript
// The TypeScript SDK (@evalguard/sdk) exposes TYPED methods — runEval,
// getEval, runSecurityScan, checkFirewall, … — not a generic request().
// For an arbitrary endpoint, call it directly:
const res = await fetch("https://evalguard.ai/api/v1/security/auto-attack", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.EVALGUARD_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
"projectId": "00000000-0000-0000-0000-000000000000",
"targetModel": "string",
"goal": "<Attack goal; required unless useLibrary=>",
"maxRounds": 1,
"candidatesPerRound": 1,
"categories": [
"string"
],
"strategy": "tap",
"targetSystemPrompt": "string",
"stopOnFirstSuccess": false,
"useLibrary": false,
"libraryCategory": "string",
"libraryCount": 1
}),
});
console.log(res.status, await res.json());Python
# The Python SDK (pip install evalguardai) exposes TYPED methods on
# EvalGuardClient — run_eval, get_eval, … — not a generic request().
# For an arbitrary endpoint, call it directly:
import os
import requests
headers = {"Authorization": f"Bearer {os.environ['EVALGUARD_API_KEY']}"}
headers["Content-Type"] = "application/json"
response = requests.request(
"POST",
"https://evalguard.ai/api/v1/security/auto-attack",
headers=headers,
json={
"projectId": "00000000-0000-0000-0000-000000000000",
"targetModel": "string",
"goal": "<Attack goal; required unless useLibrary=>",
"maxRounds": 1,
"candidatesPerRound": 1,
"categories": [
"string"
],
"strategy": "tap",
"targetSystemPrompt": "string",
"stopOnFirstSuccess": False,
"useLibrary": False,
"libraryCategory": "string",
"libraryCount": 1
},
)
print(response.status_code, response.json())Go
package main
import (
"context"
"fmt"
"net/http"
"os"
"strings"
)
func main() {
body := strings.NewReader(`{"projectId":"00000000-0000-0000-0000-000000000000","targetModel":"string","goal":"<Attack goal; required unless useLibrary=>","maxRounds":1,"candidatesPerRound":1,"categories":["string"],"strategy":"tap","targetSystemPrompt":"string","stopOnFirstSuccess":false,"useLibrary":false,"libraryCategory":"string","libraryCount":1}`)
req, _ := http.NewRequestWithContext(context.Background(), "POST", "https://evalguard.ai/api/v1/security/auto-attack", body)
req.Header.Set("Authorization", "Bearer "+os.Getenv("EVALGUARD_API_KEY"))
req.Header.Set("Content-Type", "application/json")
resp, err := http.DefaultClient.Do(req)
if err != nil { panic(err) }
defer resp.Body.Close()
fmt.Println(resp.Status)
}Errors
400401403429500