Please update the 1password MCP Server documentation to include instructions for Claude Desktop, Claude Code and Opencode
The MCP server docs has two named tabs: Codex and Kiro. Everything else — including Claude Code or Claude Desktop — is "Other."
I am trying to understand if the MCP feature is only for desktop applications or if it can be used with applications that run in the terminal.
The 1Password desktop app is the MCP server.
Codex, Kiro, Opencode, Claude Desktop and Claude Code are MCP clients (aka coding agents, or what some developers call harnesses) the wrapper of tools, permissions, and memory around the model.
Do MCP Clients need to be part of the trusted access layer? Is that what the OpenAI announcement is about?
Market reality (source): Claude Code holds the highest published SWE-bench Verified score (88.6%) and is rated the deepest programmable harness in the field.
OpenCode: 171K GitHub stars, 1.68M weekly downloads. Both are "Other."
Codex has 5M weekly users
Kiro — zero published adoption data, launched July 2025 to replace end-of-support Amazon Q — gets named placement.
- Add a Claude Code tab.
- Add an OpenCode tab.
- Document the known issues in those tabs: rate-limit behavior, biometric prompt batching, non-login shell token inheritance.
