skills.rest
ruvnet
ruvnetCOMMUNITY
Explore all skills by this author →
Website𝕏@ruvGitHub

managed-agent

Create and control Anthropic Claude Managed Agent cloud sessions with containers and tools.

70.1k|8.4k|Updated Jun 2, 2025
Search Tags:#mcp#container#anthropic#CLAude#managed agent#cloud agent#Session Management
One-click install
npx skills add https://github.com/ruvnet/claude-flow --skill managed-agent
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: managed-agent
Source: https://github.com/ruvnet/claude-flow/tree/main/plugins/ruflo-agent/skills/managed-agent
Command: npx skills add https://github.com/ruvnet/claude-flow --skill managed-agent

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Running long-lived or resource-heavy agent tasks locally is limited by your machine; this Skill provisions Anthropic Claude Managed Agents in the cloud with a real container, pre-installed packages, network access, and a persistent filesystem and transcript across turns.

Core Features & Use Cases

  • Session Lifecycle Management: Create cloud agent sessions with a chosen model, system prompt, packages (pip, npm, apt, cargo, gem, go), MCP servers, and skills, then prompt, inspect, list, and terminate them.
  • Event & Transcript Inspection: Poll session status and retrieve the full event log including user turns, agent thinking, tool_use, and tool_result entries.
  • Cost & Resource Hygiene: List all org sessions and terminate stale ones to stop billing for container time and tokens.
  • Use Case: Offload a multi-hour data processing job to a cloud container with specific pip packages installed, prompt it, monitor progress via events, and terminate the session when finished.

Quick Start

Create a managed agent session with the haiku model, send it a prompt to run a shell command in its container, then terminate the session when it finishes.

Frequently Asked Questions about managed-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a Claude Managed Agent cloud session?▼

Call managed_agent_create with optional model, system prompt, name, networking, packages, initScript, mcpServers, and skills. It returns a sessionId, agentId, and environmentId that you use for all subsequent prompt, status, and terminate calls.

When should I use a managed cloud agent instead of a local WASM agent?▼

Use the managed runtime for long-running tasks lasting minutes to hours that need a real cloud container with pre-installed packages, network access, and a persistent filesystem. Use the local WASM runtime for fast, free, ephemeral, offline work with untrusted code.

Does the managed agent require an API key?▼

Yes, it requires ANTHROPIC_API_KEY or CLAUDE_API_KEY with Claude Managed Agents beta access. Without a key, every managed_agent tool returns a structured error directing you to the local wasm_agent_create runtime instead.

Can a cloud managed agent use my local MCP server?▼

No, a local ruflo mcp start server is not reachable from Anthropic's cloud. You must deploy or tunnel an HTTP-accessible ruflo MCP server and pass its URL in the mcpServers configuration for the cloud agent to use those tools.

Why should I terminate managed agent sessions when done?▼

Cloud sessions keep billing container time and tokens until deleted. Always call managed_agent_terminate with the sessionId and environmentId, and use managed_agent_list to find and clean up stale orphaned sessions.

Related Skills

Explore More Software Engineering Skills→
AgentPMTAgentPMT

install-agentpmt-mcp

Install and configure the AgentPMT MCP server for AI clients.

Official
Intermediate
BelichickGillisMuskBelichickGillisMusk

tps-report

Monitor task progress and costs within AI agent workflows.

Official
Intermediate
reggiechan74reggiechan74
5

administration

Configure OpenTelemetry exporters to monitor Claude Code usage and costs.

Community
Advanced
ollieb89ollieb89

openclaw-sandbox

Manage Docker sandboxing configurations for tool execution in OpenClaw.

Community
Intermediate