ObeyAIS
AGENT DIRECTORY & PLAYGROUND
Connecting large language models to local databases, standard APIs, and client-side browser interfaces. Explore and simulate agentic control protocols below.
WebMCP In-Browser Tool Sandbox
WebMCP is a W3C-community standard allowing web applications to register functions inside window.navigator.modelContext. This allows browsing agents to interact with pages semantically without scraping DOM elements. Configure a tool and execute queries in the terminal simulator below.
1. EXPOSE WEBMCP TOOL
2. SIMULATED AGENT INQUIRY
A2A Peer-to-Peer Agent Mesh Simulator
The A2A (Agent-to-Agent) protocol is an emerging standard developed by Google and IBM (hosted by the Linux Foundation) allowing independent agent nodes to perform service discovery, exchange JSON-LD capabilities cards, negotiate task parameters, and delegate sub-tasks asynchronously. Trigger a peer-to-peer negotiation sequence below.
1. AGENT NODE SETTINGS
2. PROTOCOL WIRE TRANSACTION LOGS
Model Intelligence & Protocol Compliance Matrix
Evaluating large language models on protocol compliance, context-window efficiency when handling multiple schemas, latency, and tool-call accuracy.
| Model Identifier | Tool-Call Accuracy | WebMCP Support | Average Latency | Context Efficiency |
|---|---|---|---|---|
| Claude 3.5 Sonnet | 96.2% | Full Support | 1.2s | Excellent |
| GPT-4o | 94.8% | Full Support | 0.9s | Good |
| Gemini 1.5 Pro | 93.5% | Full Support | 1.4s | Excellent (1M+ Tokens) |
| Llama 3.1 70B (Instruct) | 88.1% | Partial (JSON-RPC) | 1.1s | Moderate |
* Latency is measured as Time-to-First-Token (TTFT) across 15 iterations on standard JSON-RPC over SSE channels.
Oversight & Compliance Research Archive
Access ObeyAIS's core philosophical documents and historical case studies detailing AI agency emergence, protocol failure, and compliance auditing.