Integrations & Models
Connect Claude Code to the outside world and choose the right model for every task.
MCP (Model Context Protocol) is an open standard that gives Claude Code structured access to external tools and data. Understanding its architecture — tools, resources, prompts, transport types, and context cost — is the foundation for using it effectively.
The complete reference for adding, configuring, and verifying MCP servers in Claude Code — covering the CLI commands, JSON config format, stdio vs HTTP transport, environment variable injection, and in-session management.
Claude Code handles the full OAuth flow for services like GitHub, Slack, Linear, and Google Workspace — browser-based authentication, automatic token refresh, and clean revocation via /mcp.
MCP tools follow the same permission model as Claude Code's built-in tools. Learn the permission syntax, how to scope rules per-project vs globally, and the destructive operation audit every new server requires.
A decision guide for the most impactful MCP servers: GitHub vs gh CLI, PostgreSQL, Filesystem, Slack, Linear, Notion, Playwright, and Figma. Plus the April 2026 push notification capability that makes Claude Code event-driven.
When no existing MCP server covers your internal APIs, proprietary databases, or company-specific tools, you build your own. This article covers the minimal server structure, tool and resource handlers, and the April 2026 maxResultSizeChars feature that prevents silent truncation of large outputs.
Remote MCP servers are HTTP endpoints that your whole team connects to — deployed once, used everywhere. This article covers architecture, authentication, hosting options, and the operational concerns that separate a working prototype from a production-grade team tool.
Opus 4.7, Sonnet 4.6, Haiku 4.5 — three capability tiers with real cost differences. Understanding when each model earns its price tag is one of the highest-leverage decisions in your Claude Code workflow.
Effort levels control how many thinking tokens Claude allocates before responding. From the model selection angle, the right effort level is as important as the right model — and the two choices compound.
Run Claude through AWS Bedrock or Google Vertex AI for compliance, data residency, and existing cloud spend integration. The April 2026 interactive Bedrock setup wizard and modelOverrides feature make configuration significantly easier.
Claude Code costs compound across a team. npx ccusage, right-sizing models and effort levels, disabling unused MCP servers, and the --max-budget-usd flag are the levers you have.
Every agent definition should explicitly set its model. Without it, sub-agents inherit the parent session's model — and 30 Opus sub-agents running simultaneously is an expensive accident.