cpp-load-mcp-docs git:20260822.e3a8d80B

Immutable. This exact content is served forever at /api/v1/blob/fba7cd725ead92d2.

---
name: "cpp-load-mcp-docs"
description: "Load MCP Second Opinion server documentation"
---
<!-- GENERATED by claude-power-pack - scripts/codex-skill-sync.py; edit .claude/commands/cpp/load-mcp-docs.md instead -->

## Codex harness adaptations

Generated from a Claude Code command. Where the procedure references these Claude-only surfaces, adapt as follows:

- MCP tools: use the MCP servers configured in `~/.codex/config.toml`, or fall back to the referenced repo scripts and CLI entry points.

The MCP Second Opinion server is external: it lives in the `cooneycw/mcp-second-opinion`
repo and is connected to this project through the root `.mcp.json` streamable-http pointer
(`${SECOND_OPINION_URL:-http://127.0.0.1:8080}/mcp` - default localhost 8080, per-host
override via the `SECOND_OPINION_URL` env var, issue #633). Summarize the available tools
and usage patterns from what is reachable here:

1. Read: `README.md` (MCP Second Opinion section) for the setup and connection model.
2. Inspect the live `mcp__second-opinion__*` tool definitions exposed by the connected
   server (e.g. `list_available_models`, `get_code_second_opinion`,
   `get_multi_model_second_opinion`, `create_session`, `consult`). For full server
   internals, see the external `cooneycw/mcp-second-opinion` repository.

Provide an overview of:
- The 10 available tools and their purposes
- Key features (Gemini 3 Pro, multi-turn sessions, agentic tool use)
- Cost tracking and limits
- Security features (SSRF protection, domain approval)
- Best practices for using the MCP server
- Playwright integration for web debugging

## Tavily MCP Server

The Tavily MCP server is the upstream [tavily-ai/tavily-mcp](https://github.com/tavily-ai/tavily-mcp)
package, registered via npx/stdio by `/cpp:init`. It provides four tools:

- `tavily-search` - real-time web search (basic/advanced/fast/ultra-fast depth, topic filters, date ranges)
- `tavily-extract` - extract content from URLs (markdown or text format)
- `tavily-crawl` - graph-based website traversal with intelligent discovery
- `tavily-map` - generate comprehensive sitemaps through intelligent traversal

API key (`TAVILY_API_KEY`) stored in AWS Secrets Manager `claude-power-pack/mcp-keys`.
Requires Node.js 20+. Registration: `claude mcp add tavily --transport stdio --scope user -e TAVILY_API_KEY=... -- npx -y tavily-mcp@latest`

Format the response as a concise reference guide.