AGENTS.md — RunAgents Agent Project¶

Project¶

This project contains AI agent code that deploys to the RunAgents platform. RunAgents orchestrates agents with secure, policy-driven access to external tools and services.

Capabilities¶

Capability	Description
Deploy agents	Upload Python source code; the platform analyzes, builds, and deploys
Register tools	Define external APIs (Stripe, GitHub, Slack, etc.) with auth and access control
Configure LLM providers	Route model calls through a unified gateway (OpenAI, Anthropic, Bedrock)
Monitor runs	Track agent invocations, events, and tool calls
Approve access	Just-in-time approval workflow for sensitive tool access

Tools¶

RunAgents CLI (`runagents`)¶

Install: npm install -g @runagents/cli or brew install runagents-io/tap/runagents

Command	Description
`runagents deploy --name NAME --file agent.py`	Deploy agent from source
`runagents analyze --file agent.py`	Preview code analysis results
`runagents agents list`	List deployed agents
`runagents tools list`	List registered tools
`runagents tools create --file tool.json`	Register a new tool
`runagents models list`	List model providers
`runagents models spend`	Inspect model spend and budget posture
`runagents agents config get NAME`	Inspect agent model budgets and usage
`runagents runs list --agent NAME`	List runs for an agent
`runagents approvals list`	List pending access requests
`runagents approvals approve ID`	Approve an access request
`runagents starter-kit`	Seed demo tools and model provider

RunAgents MCP Server (`runagents-mcp`)¶

Install: pip install runagents-mcp

Provides tools for direct platform access: list_agents, get_agent, get_agent_config, list_tools, list_models, get_model_spend, list_runs, get_run_events, export_context, analyze_code, deploy_agent, create_tool, validate_plan, apply_plan, update_agent_config, approve_request, seed_starter_kit.

Workflows¶

Deploy a new agent¶

Write agent code (Python) that calls tools via HTTP and uses the LLM gateway
Run runagents analyze --file agent.py to verify tool and LLM detection
Register any new tools with runagents tools create
Deploy with runagents deploy --name my-agent --file agent.py

Add a new tool¶

Register: runagents tools create --file stripe-tool.json
Update agent code to call the new tool URL
Re-deploy the agent

Handle approvals¶

Check: runagents approvals list
Review the request details
Approve: runagents approvals approve <request-id>

References¶

Documentation: https://docs.runagents.io
Writing Agents: https://docs.runagents.io/getting-started/writing-agents/
API Reference: https://docs.runagents.io/api/overview/
Architecture: https://docs.runagents.io/concepts/architecture/