AppOmni Agent Guard API

openapi: 3.2.0
info:
  title: AppOmni AI Agent Guard API
  description: 'Marlin AI autonomous analysis plans and the AgentGuard prompt classification endpoint.


    Derived faithfully from the AppOmni public Postman collection published by AppOmni at https://api.appomni.com/ — every path, method, parameter, request body and example response below is taken verbatim from that collection. No operation was invented.'
  version: 1.0.0
  contact:
    name: AppOmni
    url: https://appomni.com/support/
  license:
    name: Proprietary
    url: https://appomni.com/terms-of-service/
  x-generated-from: AppOmni public Postman collection (publishedId 2sBXc7Mjib) + AppOmni's published @appomni/n8n-nodes-agentguard source
  x-generated-source: https://api.appomni.com/api/collections/45135595/2sBXc7Mjib?segregateAuth=true&versionTag=latest
  x-generated-method: derived
  x-generated-date: '2026-09-04'
servers:
- url: https://{instance}.appomni.com
  description: AppOmni tenant. Replace {instance} with your AppOmni subdomain — e.g. `acme` if you log in at acme.appomni.com.
  variables:
    instance:
      default: example
      description: Your AppOmni tenant subdomain
security:
- bearerAuth: []
tags:
- name: AgentGuard
  description: AppOmni AgentGuard runtime prompt security — DLP and prompt-firewall classification for AI agent traffic.
paths:
  /api/v1/ai/prompts/agents/classify:
    post:
      operationId: classifyAgentPrompt
      summary: Classify an agent prompt (AgentGuard)
      tags:
      - AgentGuard
      description: 'Send a prompt to AppOmni AgentGuard for DLP and/or prompt-firewall analysis. Returns an allow/block verdict with per-classifier scores and block reasons. Authenticated with an AppOmni-issued ingest token in the `X-AppOmni-Ingest-Token` header (not the platform bearer token).


        Schema derived verbatim from AppOmni''s own published n8n community node source (github.com/appomni/n8n-nodes-agentguard, credentials/AgentGuardApi.credentials.ts and nodes/AgentGuard/Agentguard.node.ts).'
      security:
      - ingestToken: []
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/AgentGuardClassifyRequest'
            example:
              messages:
              - role: user
                content: AppOmni AgentGuard credential test
                content_type: text/plain
              include_details: false
      responses:
        '200':
          description: Classification verdict
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/AgentGuardClassifyResponse'
        '401':
          description: Unauthorized — missing or invalid ingest token
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '403':
          description: Forbidden
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
components:
  schemas:
    AgentGuardClassifierResult:
      type: object
      title: AgentGuardClassifierResult
      properties:
        name:
          type: string
        type:
          type: string
        status:
          type: string
          enum:
          - success
          - timeout
          - error
          - skipped
        duration_ms:
          type: integer
        applied_threshold:
          type: number
        scores:
          type: object
          properties:
            benign:
              type: number
            malicious:
              type: number
        block_reasons:
          type: array
          items:
            type: string
        details:
          type: object
          additionalProperties: true
        error:
          type: string
        skip_reason:
          type: string
    AgentGuardClassifyRequest:
      type: object
      title: AgentGuardClassifyRequest
      required:
      - messages
      properties:
        messages:
          type: array
          items:
            $ref: '#/components/schemas/AgentGuardPrompt'
        include_details:
          type: boolean
          description: Return per-classifier detail in the response
        metadata:
          type: object
          title: AgentGuardClassifyMetadata
          description: Optional caller context used for attribution and user quarantine
          properties:
            user:
              type: object
              properties:
                id:
                  type: string
                username:
                  type: string
                email:
                  type: string
                principal_type:
                  type: string
            session:
              type: object
              properties:
                id:
                  type: string
            agent:
              type: object
              properties:
                id:
                  type: string
                name:
                  type: string
            request:
              type: object
              properties:
                src_app:
                  type: string
                interface:
                  type: string
                user_agent:
                  type: string
                src_ip:
                  type: string
    AgentGuardClassifyResponse:
      type: object
      title: AgentGuardClassifyResponse
      required:
      - response_action
      - response_message
      - scores
      - block_reasons
      - classifiers
      - effective_threshold
      properties:
        response_action:
          type: string
          enum:
          - allow
          - block
          description: Verdict — `block` means the prompt must not be forwarded to the model
        response_message:
          type: string
        scores:
          type: object
          properties:
            benign:
              type: number
            malicious:
              type: number
        block_reasons:
          type: array
          items:
            type: string
        classifiers:
          type: array
          items:
            $ref: '#/components/schemas/AgentGuardClassifierResult'
        effective_threshold:
          type: number
        event_id:
          type: string
    Error:
      type: object
      title: Error
      description: Standard Django REST Framework error envelope returned by the AppOmni API.
      properties:
        detail:
          type: string
          description: Human readable error message
    AgentGuardPrompt:
      type: object
      title: AgentGuardPrompt
      required:
      - role
      - content
      - content_type
      properties:
        role:
          type: string
          description: Message role, e.g. `user` or `system`
        content:
          type: string
          description: The prompt text to classify
        content_type:
          type: string
          description: Media type of `content`
          example: text/plain
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: 'AppOmni API access token, created and managed in the AppOmni platform under Settings > API Settings. Sent as `Authorization: Bearer <token>`.'
    ingestToken:
      type: apiKey
      in: header
      name: X-AppOmni-Ingest-Token
      description: AppOmni-issued ingest token used by AgentGuard and the AODP ingest endpoint. Sent as the `X-AppOmni-Ingest-Token` request header.