← All notes

Answer: An MCP server exposes tools and resources to an agent over a transport (stdio, SSE, or Streamable HTTP). Filter tools per server, and bypass tool call confirmations only for servers you trust.

Transports

• Stdio: spawns a subprocess and communicates via stdin/stdout.
• SSE: connects to Server-Sent Events endpoints.
• Streamable HTTP: uses HTTP streaming.

The OpenAI Agents SDK maps these to MCPServerStdio, MCPServerSse, and MCPServerStreamableHttp, plus hosted MCP server tools.

Gemini CLI per-server settings (mcpServers in settings.json)

• includeTools is an allowlist; excludeTools takes precedence over includeTools.
• trust: true bypasses all tool call confirmations for that server (default false).
• timeout defaults to 600,000ms (10 minutes).

Trust before connecting

MCP tools may expose model-context data and act using your credentials. The SDK docs advise connecting only to trusted servers, giving least-privilege credentials, and requiring approval for sensitive operations.

Artifacts

• MCP transports and trust surfaces (table)
• MCP transports and trust surfaces (markdown table)
• Topic packet: ao-mcp

See also

• ao-mcp
• ao-security-gemini-cli-sandbox

Claims

• id: C1 claim: MCP servers expose tools and resources. source: S1 evidence: "exposes tools and resources to the Gemini CLI through the Model Context Protocol"

• id: C2 claim: excludeTools beats includeTools. source: S1 evidence: "excludeTools always takes precedence over includeTools"

• id: C3 claim: trust true bypasses confirmations. source: S1 evidence: "When true, bypasses all tool call confirmations for this server"

• id: C4 claim: Connect only to trusted servers. source: S2 evidence: "Connect only to servers you trust, use least-privilege credentials"

Sources

• [S1] MCP servers with Gemini CLI. Google Gemini CLI docs (GitHub). https://github.com/google-gemini/gemini-cli/blob/main/docs/tools/mcp-server.md . License: Apache-2.0 (google-gemini/gemini-cli LICENSE). Accessed 2026-11-09.

• [S2] Model context protocol (MCP). OpenAI Agents SDK (Python) docs (GitHub). https://github.com/openai/openai-agents-python/blob/main/docs/mcp.md . License: MIT (openai/openai-agents-python LICENSE). Accessed 2026-11-09.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted). Sources: GitHub repository docs under Apache License 2.0 (https://www.apache.org/licenses/LICENSE-2.0) or MIT License (https://opensource.org/licenses/MIT), as listed per source; copyright notices remain with the original authors. Text was condensed and changed from the originals.

@misc{iil_note_qn3d4nq7,
  author = {{iinventedlight}},
  title = {{MCP servers in agent CLIs: transports, tool filters, trust}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-09},
  url = {https://iinventedlight.com/library/notes/ao-mcp-transports-and-trust/},
  note = {Permanent ID iil:note:qn3d4nq7. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/ao-mcp-transports-and-trust.md.