Okareo

Name: Okareo
Author: okareo-ai

@okareo-ai

关于 Okareo

Simulate and evaluate voice & text agents from your editor. Generate synthetic-user "Drivers" from your codebase, run multi-turn simulations against your agents, and pull back transcripts, traces, and scores without leaving your AI assistant.

#agent-simulation #evaluation #testing #voice-agents #ai-agents #chat-agents #voice #chat #observability #llm #qa #ci #synthetic-data

基本信息

分类

其他

传输方式

stdio

发布者

okareo-ai

提交者

Matt Wyman

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "okareo": {
      "type": "http",
      "url": "https://tools.okareo.com/mcp"
    }
  }
}

工具

** `save_scenario`, `list_scenarios`, `get_scenario`, `create_scenario_version`

** `create_or_update_driver`, `list_driver_voices`, `get_driver`, `list_drivers`

** `create_or_update_target`, `get_target`, `list_targets`

** `run_simulation`, `list_simulations`

** `create_or_update_check`, `generate_check`

** `ingest_conversations`, `connect_voice_integration`, `get_voice_webhook_url`

概览

What is Okareo?

Okareo is an MCP server for agent simulation and evaluation of voice and text agents. It enables developers to run dozens of synthetic conversations against their agent directly from MCP-compatible editors like Claude Code, Cursor, or any MCP-ready tool, without installing local infrastructure.

How to use Okareo?

Point your MCP editor at the hosted endpoint https://tools.okareo.com/mcp. For Claude Code, run claude mcp add --transport http okareo https://tools.okareo.com/mcp. For any MCP client, add the server URL to your configuration JSON. Authentication uses OAuth on first connect; if OAuth discovery is not supported, set the OKAREO_API_KEY environment variable for a Bearer-header fallback.

Key features of Okareo

Generate synthetic users (Drivers) from your agent's system prompt and tools.
Run voice and text simulations in parallel across multiple languages.
Replay production failures as reusable test Drivers.
Run judge, symbolic, and audio-based evaluation checks.
Re-score past runs without re-running the model.
Query analytics and pin dashboards.

Use cases of Okareo

Stress-test a multilingual voice agent with dozens of synthetic callers.
Convert real failed calls into automated regression tests.
Compare model performance across different LLMs in multi-turn conversations.
Evaluate policy compliance by generating boundary-pushing Drivers from agent code.

FAQ from Okareo

What does Okareo MCP do?

It turns your AI coding assistant into an evaluation workbench — you describe tests in natural language, and the server builds synthetic users, runs simulations, and returns transcripts, traces, and scores.

Do I need to install anything locally?

No. Okareo is hosted — there is nothing to install; just point your editor at the endpoint and sign in once through your browser.

Which editors and MCP clients are supported?

Claude Code, Claude Desktop, Cursor, VS Code 1.101+, Cline, Gemini Code Assist, GitHub Copilot, Windsurf, and any client that speaks MCP's streamable-HTTP transport.

How does authentication work?

OAuth on first connect (a browser tab opens to app.okareo.com). If your client does not implement MCP OAuth discovery, set OKAREO_API_KEY in your environment as a Bearer-header fallback.