feat(mcp): unified MCP server registration for tool and resource access (RFC-0058)

examples/kb_data/docs/architecture.md

@@ -0,0 +1,17 @@
+# Architecture Overview
+
+AgentPool bridges multiple protocols with native PydanticAI agents.
+
+## Layers
+
+1. **Configuration** — YAML manifests parsed into Pydantic models
+2. **Orchestration** — EventBus, SessionController, RunLoop
+3. **Protocols** — ACP, OpenCode, MCP, AG-UI, OpenAI API
+4. **Capabilities** — Tools, skills, MCP servers, resources
+
+## Message Flow
+
+```
+Client → Protocol Server → SessionController → Agent (PydanticAI)
+    → Tool Execution → Event Bus → Protocol Converter → Client
+```

examples/kb_data/docs/intro.md

@@ -0,0 +1,16 @@
+# Introduction to AgentPool
+
+AgentPool is a unified agent orchestration framework for
+YAML-based configuration of heterogeneous AI agents.
+
+## Core Philosophy
+
+Define once in YAML, expose through multiple protocols,
+enable seamless inter-agent collaboration.
+
+## Key Features
+
+- Multi-agent orchestration (teams, chains)
+- ACP, OpenCode, MCP, AG-UI protocol support
+- YAML-based agent definition
+- Structured output with Pydantic

examples/kb_data/docs/quickstart.md

@@ -0,0 +1,23 @@
+# Quick Start
+
+## Installation
+
+```bash
+uv tool install wolfharness
+```
+
+## Minimal config
+
+```yaml
+agents:
+  assistant:
+    type: native
+    model: openai:gpt-4o
+    system_prompt: "You are a helpful assistant."
+```
+
+## Run
+
+```bash
+wolfharness run assistant "Hello!"
+```

examples/kb_data/docs/readme.md

@@ -0,0 +1,3 @@
+# Docs subdirectory
+
+A nested markdown file to exercise the `file://{path}` template with a subdirectory.