@perssona/orchestrator
MCP server that lets AI agents control Perssona. It enables creating terminals, managing canvas nodes, sending prompts, and running multi-agent workflows.
README
@perssona/orchestrator
MCP server that lets AI agents control Perssona — create terminals, manage canvas nodes, send prompts, and run multi-agent workflows.
Requirements
- Perssona must be running — the MCP communicates with the local bridge on
127.0.0.1:17381 - Node.js 18+
- Claude Code or Codex CLI
Installation
Claude Code
claude mcp add --scope user perssona-orchestrator -- npx -y @perssona/orchestrator
Codex
Add to ~/.codex/config.toml:
[mcp_servers.perssona-orchestrator]
command = "npx"
args = ["-y", "@perssona/orchestrator"]
startup_timeout_sec = 60
Available Tools
| Tool | Description |
|---|---|
create_connected_terminal |
Create a terminal node connected to a source node |
create_codex_terminal_from_agent |
Create a Codex terminal from an agent preset |
create_claude_terminal_from_agent |
Create a Claude terminal from an agent preset |
create_agent_pair |
Create two connected LLM terminals for parallel roles |
run_workflow_from_annotation |
Run a multi-agent workflow from an annotation or prompt |
send_text_to_terminal |
Send text directly to a specific terminal node |
send_text_to_connected_terminals |
Broadcast text to all terminals connected to a node |
send_annotation_to_connected_terminals |
Send annotation content to connected terminals |
create_annotation_node |
Create an annotation node on the canvas |
update_annotation_node |
Update an existing annotation node |
connect_nodes |
Connect two canvas nodes with an edge |
disconnect_nodes |
Remove the edge between two nodes |
remove_node |
Remove a node and its edges |
focus_node |
Select and center the viewport on a node |
add_node_to_grid |
Add a node to the grid view |
open_nodes_in_grid |
Open one or more nodes in the grid view |
get_selected_node |
Return the currently selected canvas node |
get_connected_nodes |
Return nodes connected to a given node |
list_canvas_nodes |
List all nodes and edges in the active canvas |
list_orchestrator_presets |
List workspace presets (models, agents, workflows) |
Debug
Set PERSSONA_MCP_DEBUG=1 to write a log to your system temp directory:
PERSSONA_MCP_DEBUG=1 node server.mjs
Contributing
Contributions are welcome via Pull Requests.
- The
mainbranch is protected — direct pushes are not allowed - Open an issue first for bugs or feature proposals
- Keep PRs focused and well-described
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。