Guides · Claude Code
Configure Claude Code with MCP servers, project rules, and hooks for autonomous AI development.
Claude Code by Anthropic is a terminal-based AI coding agent that runs locally. It supports MCP (Model Context Protocol) for tool integration, project-level rules via .mcp.json, and autonomous execution through hooks. AACFlow provides ready-to-use MCP servers for your workflows, knowledge bases, and data pipelines.
1. Get your AACFlow API key from Dashboard → Settings → API Keys. 2. Run: claude mcp add aacflow --scope project -- npx -y @aacflow/mcp-server 3. Set the environment variable: AACFLOW_API_KEY=ak_xxxx 4. Verify with: claude mcp list (should show aacflow as connected)
Claude Code supports three MCP scopes: local (per-project, private), project (.mcp.json in repo root, team-shared), and user (global across all projects). Add AACFlow's MCP server to your project: claude mcp add aacflow --scope project --transport http https://mcp.aacflow.io For team setups, commit .mcp.json to Git. Each developer supplies their own API key via environment variable.
1. Create a .claude/settings.json with your AACFlow project ID. 2. Define a hook in .claude/hooks.json to trigger AACFlow workflows on file save. 3. Run: claude "Check my AACFlow knowledge base for the error in src/api.ts and propose a fix" 4. Claude will search your KB, retrieve context, and suggest code changes with full traceability.
• Use project-scoped MCP for team workflows, user-scoped for personal tools. • Set CLAUDE_PROJECT_DIR for environment-aware configs. • Keep .mcp.json in version control for team consistency. • Use hooks for pre-commit checks and automated testing. • Rotate API keys regularly and never commit them to Git. • For large codebases, configure .claudeignore to exclude node_modules and build artifacts.
