openapi: 3.0.3
info:
  title: Agno AgentOS Agents Evals API
  description: 'AgentOS is Agno''s self-hostable runtime and API server. It turns agents, teams, and workflows defined with the open-source Agno Python framework into a RESTful API exposing runs, sessions, memory, knowledge, and evals. There is no single Agno-operated multi-tenant base URL - every AgentOS instance runs on infrastructure you control, defaulting to http://localhost:7777 in local development. When a Security Key (OS_SECURITY_KEY) is configured, all routes require an "Authorization: Bearer <token>" header; otherwise authentication is disabled. This document models the endpoints described in Agno''s public API reference (docs.agno.com/reference-api) plus the sibling endpoints named in Agno''s own documentation index (docs.agno.com/llms.txt). Some request/response field names are reconstructed from that reference content rather than a downloaded machine-readable OpenAPI file, so treat schemas as representative rather than byte-exact.'
  version: '1.0'
  contact:
    name: Agno
    url: https://www.agno.com
  license:
    name: Mozilla Public License 2.0
    url: https://github.com/agno-agi/agno/blob/main/LICENSE
servers:
- url: http://localhost:7777
  description: Local AgentOS (default self-hosted address)
security:
- bearerAuth: []
- {}
tags:
- name: Evals
  description: Evaluation runs for quality and reliability.
paths:
  /eval-runs:
    get:
      operationId: listEvaluationRuns
      tags:
      - Evals
      summary: List evaluation runs
      responses:
        '200':
          description: A list of evaluation runs.
          content:
            application/json:
              schema:
                type: array
                items:
                  $ref: '#/components/schemas/EvalRun'
    post:
      operationId: executeEvaluation
      tags:
      - Evals
      summary: Execute an evaluation
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/EvalRunInput'
      responses:
        '200':
          description: The created evaluation run.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/EvalRun'
components:
  schemas:
    EvalRun:
      type: object
      properties:
        eval_run_id:
          type: string
        eval_type:
          type: string
        status:
          type: string
        score:
          type: number
    EvalRunInput:
      type: object
      required:
      - eval_type
      - input
      properties:
        eval_type:
          type: string
          enum:
          - accuracy
          - agent_as_judge
          - performance
          - reliability
        input:
          type: string
          minLength: 1
        agent_id:
          type: string
        team_id:
          type: string
        model_id:
          type: string
        model_provider:
          type: string
        expected_output:
          type: string
        criteria:
          type: string
        expected_tool_calls:
          type: array
          items:
            type: string
        scoring_strategy:
          type: string
          enum:
          - numeric
          - binary
          default: binary
        threshold:
          type: integer
          minimum: 1
          maximum: 10
          default: 7
        num_iterations:
          type: integer
          minimum: 1
          maximum: 100
          default: 1
        warmup_runs:
          type: integer
          minimum: 0
          maximum: 10
          default: 0
        name:
          type: string
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: HTTP Bearer token equal to OS_SECURITY_KEY. Required only when a Security Key is configured on the AgentOS instance; disabled otherwise.