docs: update MCP and OpenAI configuration guides

[cursor]

Summary

Updated administrator and user documentation to match the shipped configuration changes from PR #617:

• Documented that the MCP client and embedded Mattermost MCP server are always enabled, and that admins now manage per-tool policies from the Tools tab.
• Documented that direct OpenAI services always use the Responses API, while OpenAI Compatible and Azure retain the Use Responses API setting.
• Clarified bot capability availability and defaults for native web search, reasoning, and structured output.
• Removed stale MCP setup guidance and a stale screenshot reference tied to removed toggles.

Documentation files updated:

• docs/admin_guide.md - updated service, bot, built-in web search, MCP configuration, and Mattermost MCP server sections to match the current System Console behavior.
• docs/providers.md - documented Responses API behavior for OpenAI, OpenAI Compatible, and Azure, and corrected the Azure service-type selection step.
• docs/user_guide.md - updated MCP tools wording so it no longer implies a global MCP enable toggle.

A few clarifications were informed by PR context from #617 and verified against the merged code:

• The recommendation to use OpenAI Compatible instead of direct OpenAI when legacy Chat Completions compatibility is required.
• The emphasis that admins should use the MCP Tools tab to disable or change policy for individual tools.

No screenshots were added. The planned screenshot value was low because this change primarily removes or renames existing controls, and the existing MCP screenshot in the docs had become stale and was removed instead.

No borderline documentation-impact items were included beyond these admin-facing configuration changes.

Ticket Link

N/A

Screenshots

None.

Release Note

Updated the Agents documentation to reflect that MCP and the embedded Mattermost MCP server are always enabled, direct OpenAI services always use the Responses API, and OpenAI Compatible/Azure services retain the Use Responses API setting.

  

[github-actions]

🤖 LLM Evaluation Results

OpenAI

✅ Overall: 19/19 tests passed (100.0%)

Provider	Total	Passed	Failed	Pass Rate
✅ OPENAI	19	19	0	100.0%

---

Anthropic

✅ Overall: 19/19 tests passed (100.0%)

Provider	Total	Passed	Failed	Pass Rate
✅ ANTHROPIC	19	19	0	100.0%

---

This comment was automatically generated by the eval CI pipeline.