Skip to content

AI Assistant Setup

Set up your AI coding assistant to deploy and manage RunAgents agents directly from your editor. This guide covers template files for project context and the MCP server for live platform access.

Template Files

Template files give your AI assistant context about RunAgents commands, agent code patterns, and deployment workflows. Copy the right file into your agent project's root directory.

Claude Code — CLAUDE.md

curl -o CLAUDE.md https://docs.runagents.io/cli/templates/CLAUDE.md

Place CLAUDE.md in your project root. Claude Code reads it automatically and learns the RunAgents CLI commands, agent code patterns, and deployment workflow.

Cursor — .cursorrules

curl -o .cursorrules https://docs.runagents.io/cli/templates/cursorrules.md

Place .cursorrules in your project root. Cursor reads it automatically.

Universal — AGENTS.md

curl -o AGENTS.md https://docs.runagents.io/cli/templates/AGENTS.md

AGENTS.md is an emerging standard for describing project capabilities and tools. Works with any assistant that reads project-level markdown files.

MCP Server

The RunAgents MCP server gives your AI assistant direct access to the platform API — listing agents, deploying code, managing tools, and handling approvals — all without leaving your editor.

Install

pip install runagents[mcp]   # recommended — includes full SDK + CLI
# or
pip install runagents-mcp    # standalone MCP server only

Configure for Claude Code

Add to your Claude Code settings.json (or project .mcp.json):

{
  "mcpServers": {
    "runagents": {
      "command": "runagents-mcp",
      "env": {
        "RUNAGENTS_ENDPOINT": "https://YOUR_WORKSPACE.try.runagents.io",
        "RUNAGENTS_API_KEY": "YOUR_API_KEY",
        "RUNAGENTS_NAMESPACE": "default"
      }
    }
  }
}

Configure for Cursor

Add to your Cursor MCP settings (.cursor/mcp.json):

{
  "mcpServers": {
    "runagents": {
      "command": "runagents-mcp",
      "env": {
        "RUNAGENTS_ENDPOINT": "https://YOUR_WORKSPACE.try.runagents.io",
        "RUNAGENTS_API_KEY": "YOUR_API_KEY",
        "RUNAGENTS_NAMESPACE": "default"
      }
    }
  }
}

Available Tools

The MCP server exposes 14 tools:

Tool Description Type
list_agents List all deployed agents Read
get_agent Get agent details Read
list_tools List registered tools Read
list_models List model providers Read
list_runs List agent runs Read
get_run_events Get events for a run Read
export_context Export full workspace context Read
analyze_code Analyze source code for tools and LLM usage Read
deploy_agent Deploy an agent from source or image Mutate
create_tool Register a new tool Mutate
validate_plan Validate an action plan Read
apply_plan Apply an action plan Mutate
approve_request Approve a pending access request Mutate
seed_starter_kit Create demo tools and model provider Mutate

Configuration

The MCP server reads configuration from environment variables or ~/.runagents/config.json (the same config file used by the CLI):

Source Variable Description
Env RUNAGENTS_ENDPOINT Platform API URL
Env RUNAGENTS_API_KEY API key or workspace key (ra_ws_...)
Env RUNAGENTS_NAMESPACE Target namespace (default: default)
File ~/.runagents/config.json Fallback — shared with CLI

Environment variables take precedence over the config file.

What You Can Do

With the template file and MCP server configured, your AI assistant can:

  • Deploy agents: "Deploy agent.py as payment-agent with the stripe-api tool"
  • Register tools: "Register the Stripe API as a tool with API key auth"
  • Monitor runs: "Show me the latest runs for payment-agent"
  • Handle approvals: "List pending approvals and approve the one for stripe-api"
  • Analyze code: "Analyze agent.py and tell me what tools it calls"
  • Export context: "Show me all agents, tools, and model providers in my workspace"

Next Steps