PactKit integrates with Model Context Protocol servers to extend agent capabilities. All MCP instructions are conditional — if a server isn't available, it's gracefully skipped.

Supported Servers

Context7

Fetch up-to-date library documentation and code examples.

• Tools: resolve-library-id → get-library-docs
• When: Implementing with an unfamiliar library API
• Phase: Act

shadcn

Search, browse, and install UI components from shadcn registries.

• Tools: search_items_in_registries, view_items_in_registries, get_add_command_for_items
• When: Project has a components.json file (shadcn is configured)
• Phase: Design

Playwright MCP

Browser automation for testing — snapshots, clicks, screenshots, form filling.

• Tools: browser_navigate, browser_snapshot, browser_click, browser_take_screenshot
• When: Running browser-level QA checks
• Phase: Check

Chrome DevTools MCP

Performance tracing, console message inspection, network request analysis.

• Tools: performance_start_trace, list_console_messages, list_network_requests
• When: Performance or runtime diagnostics needed
• Phase: Check

Memory MCP

Persistent knowledge graph for cross-session context.

• Tools: create_entities, create_relations, add_observations, search_nodes
• When: Storing or retrieving architectural decisions across sessions
• Phase: Plan, Act, Done

Usage by PDCA Phase

Conditional Behavior

Each MCP integration checks for tool availability at runtime:

IF mcp__memory__create_entities tool is available:
  → store design context
ELSE:
  → skip gracefully, continue workflow

No MCP server is required for PactKit to function. They enhance the workflow when present but are never a hard dependency.