---
title: "AgentWire — Your Agent as a URL"
description: "Turn system personas, remote Model Context Protocol (MCP) connectors, and SKILL.md rules into curlable, A2A-compliant, and OpenAI-compatible serverless agent endpoints."
url: "http://127.0.0.1:9160/"
protocol_level: "Level 5: Agent-Native"
mcp_endpoint: "http://127.0.0.1:9160/mcp"
a2a_endpoint: "http://127.0.0.1:9160/api/a2a"
openai_compatible_base: "http://127.0.0.1:9160/v1"
---

# AgentWire — Your Agent as a URL

Turn prompts, remote MCP connectors, and `SKILL.md` engineering standards into curlable serverless agent endpoints supporting **MCP Streamable HTTP**, **Google A2A v1.0**, **OpenAI Chat Completions & Responses SDKs**, **SSE Streaming**, **Structured JSON Outputs**, **Reasoning Effort**, and **Automated Evals**.

## Active Agent Fleet

- **Taste & Design Critic** (`agt_design_critic`, slug: `design-critic`): Audits UI implementations against Taste Skill principles and suggests high-craft refinements
  - Run Endpoint: `POST http://127.0.0.1:9160/api/agents/agt_design_critic/run`
  - A2A Card: `GET http://127.0.0.1:9160/api/agents/agt_design_critic/agent-card.json`
  - Eval Suite: `POST http://127.0.0.1:9160/api/agents/agt_design_critic/eval`
  - Models: `anthropic/claude-3.5-sonnet` -> `google/gemini-2.5-flash` -> `meta-llama/llama-3.3-70b-instruct`
- **GitHub Issue & PR Triager** (`agt_pr_triage`, slug: `pr-triage`): Connects to GitHub MCP to analyze bug reports, verify reproduction steps, and draft fixes
  - Run Endpoint: `POST http://127.0.0.1:9160/api/agents/agt_pr_triage/run`
  - A2A Card: `GET http://127.0.0.1:9160/api/agents/agt_pr_triage/agent-card.json`
  - Eval Suite: `POST http://127.0.0.1:9160/api/agents/agt_pr_triage/eval`
  - Models: `google/gemini-2.5-flash` -> `anthropic/claude-3.5-sonnet`

## 1. Run Any Agent (JSON or SSE Stream)

```bash
curl -X POST http://127.0.0.1:9160/api/agents/agt_design_critic/run \
  -H "Content-Type: application/json" \
  -d '{
    "task": "Audit this hero button for WCAG AAA contrast and anti-slop rules",
    "reasoningEffort": "medium",
    "stream": false
  }'
```

## 2. Native Model Context Protocol (MCP) Server

Connect Claude Desktop, Cursor, Windsurf, or any MCP client directly to `http://127.0.0.1:9160/mcp`:

```bash
curl -X POST http://127.0.0.1:9160/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}}'
```

## 3. Google Agent-to-Agent (A2A v1.0) Protocol

```bash
curl -X POST http://127.0.0.1:9160/api/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": "task_1",
    "method": "tasks/send",
    "params": {
      "id": "a2a_task_01",
      "agentId": "agt_design_critic",
      "message": {
        "role": "user",
        "parts": [{"type": "text", "text": "Audit our typography hierarchy"}]
      }
    }
  }'
```

## 4. OpenAI SDK Drop-In Compatibility (`/v1/chat/completions` & `/v1/responses`)

```bash
curl -X POST http://127.0.0.1:9160/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agt_design_critic",
    "messages": [{"role": "user", "content": "Check contrast ratio requirements"}]
  }'
```

## 5. Discovery Manifests

- API Catalog (RFC 9727): `http://127.0.0.1:9160/.well-known/api-catalog`
- MCP Server Card (SEP-1649): `http://127.0.0.1:9160/.well-known/mcp/server-card.json`
- A2A Agent Card (v1.0.0): `http://127.0.0.1:9160/.well-known/agent-card.json`
- Agent Skills Index (RFC v0.2.0): `http://127.0.0.1:9160/.well-known/agent-skills/index.json`
- Web Bot Auth JWKS (RFC 9421): `http://127.0.0.1:9160/.well-known/http-message-signatures-directory`
- OpenAPI 3.1.0 Spec: `http://127.0.0.1:9160/openapi.json`
- Auth & Rate Limits: `http://127.0.0.1:9160/auth.md`