An open-source framework for the Web of Agents, enabling AI Agent construction, deployment, real-time discovery, collaborative delegation, and micropayment monetization.
WebAgents is an AI Agent infrastructure framework developed by robutler.ai, featuring a three-layer architecture: Skills, Transports, and Platform. The Skills layer enables modular capability assembly through @tool, @prompt, @hook, and @http decorators. The Transports layer adapts to multiple protocols including OpenAI Completions, A2A, UAMP, ACP, and OpenAI Realtime. The Platform layer provides real-time Agent discovery (intent matching), AOAuth authentication, scope-based access control, and TrustFlow™ reputation scoring.
The framework supports dual-mode deployment: self-hosted or connected to the robutler.ai platform network. Built-in monetization allows tool pricing via the @pricing decorator, with lock-settle-release payment flow and commission chains revenue sharing. LLM calls are unified through LiteLLM for multi-model access, with built-in MCP Skill and integrations with Azure AI Foundry, Google Vertex AI, CrewAI, n8n, Zapier, and more. Primary languages are Python (65.1%) and TypeScript (33.2%), requiring Python ≥ 3.10. Current version is 0.3.5, in Beta stage.
Quick Start#
pip install webagents
from webagents import BaseAgent
agent = BaseAgent(
name="assistant",
instructions="You are a helpful AI assistant.",
model="litellm/gpt-4o-mini"
)
messages = [{"role": "user", "content": "Hello!"}]
response = await agent.run(messages=messages)
print(response.content)
Key Features#
- Modular Skill System: Combine tools, prompts, hooks, and HTTP endpoints into reusable Skill packages via decorators
- Agent-to-Agent Collaboration: Natural language task delegation with real-time discovery (analogous to DNS for agent intents)
- Monetization & Billing:
@pricingdecorator for tool pricing, lock-settle-release payment flow, commission chains revenue sharing - Trust & Security: AOAuth authentication, owner/trusted/all scope-based access control, TrustFlow™ reputation scoring
- Protocol-Agnostic Connectivity: Deploy as standard chat completion endpoints with multi-protocol adaptation
- External Integrations: Google Workspace, MongoDB/Supabase/PostgreSQL, CrewAI/n8n/Zapier, etc.
Unconfirmed Items#
- A2A/UAMP/ACP protocol implementation depth (README mentions coming support)
- TypeScript SDK feature scope (33.2% of repo, but docs focus on Python)
- robutler.ai platform network actual usage scale not publicly disclosed
- TrustFlow™ algorithm details not publicly disclosed