@gotza02/seq-thinking

Orchestrates multi-step reasoning chains where each step's output feeds into the next step's input, enabling structured decomposition of complex problems into sequential reasoning phases. Implements a pipeline pattern that maintains state across thinking steps and enforces execution order, allowing agents to build on previous conclusions rather than reasoning in isolation.

Coordinates multiple AI agents working in parallel or sequence toward a shared goal, managing agent lifecycle, message routing between agents, and consensus/aggregation of results. Implements a swarm pattern where agents can spawn sub-agents, communicate asynchronously, and coordinate on shared state or objectives without requiring a centralized orchestrator.

Exposes sequential thinking and swarm coordination capabilities through the Model Context Protocol (MCP), allowing any MCP-compatible client (Claude Desktop, custom applications, other agents) to invoke reasoning and coordination features as remote tools. Implements MCP server specification with proper resource handling, tool definitions, and protocol compliance.

Maintains and tracks state across sequential thinking steps, preserving intermediate conclusions, context, and reasoning artifacts between steps. Implements a state machine pattern where each thinking step is a discrete state transition, with full history preservation for debugging and auditing. Allows agents to reference previous thinking steps and build cumulative reasoning.

Enables agents to dynamically spawn child agents for subtasks and manages their complete lifecycle (creation, execution, monitoring, termination). Implements a hierarchical agent pattern where parent agents can delegate work to child agents with specific roles and constraints, and collect results asynchronously. Handles agent resource cleanup and prevents resource leaks.

Exports complete reasoning traces in structured formats (JSON, markdown, etc.) suitable for visualization, analysis, and debugging. Implements trace serialization that captures the full reasoning path including intermediate steps, decisions, and conclusions. Enables external tools to visualize reasoning as flowcharts, timelines, or decision trees.