Buzz Killington

Provides semantic search across developer documentation through the Model Context Protocol, enabling LLM agents to retrieve fact-checked answers from trusted sources without hallucination. Implements a schema-based tool registry that exposes documentation queries as callable functions within the MCP protocol, allowing agents to invoke searches during reasoning chains and receive structured results with source attribution.

Provides pre-built, fact-checked prompt templates optimized for code generation tasks, delivered through MCP as callable tools. Templates encode best practices, error patterns, and domain-specific guidance to improve LLM output quality without requiring manual prompt engineering. Agents invoke these templates as structured tools, passing context variables (language, framework, problem description) to generate contextually-appropriate prompts.

Validates code generation outputs and developer queries against trusted documentation sources, returning confidence scores and source citations. Implements a verification pipeline that cross-references generated code snippets, API usage patterns, and best practices against indexed documentation, surfacing potential inaccuracies or deprecated patterns. Results include source URLs and documentation excerpts to support human review.

Analyzes coding problems by decomposing them into sub-problems and retrieving relevant documentation for each component, enabling agents to reason through complex issues with fact-checked context. Implements a multi-step analysis pipeline that identifies problem categories, retrieves applicable documentation, and synthesizes solutions grounded in trusted sources. Results include problem decomposition, relevant documentation sections, and reasoning traces.

Routes documentation queries to language and framework-specific indices, ensuring agents retrieve documentation relevant to their current development context. Implements context-aware routing that identifies the programming language, framework, and domain from query context or explicit parameters, then queries the appropriate documentation subset. Supports polyglot development workflows where agents work across multiple languages and frameworks.

Enables agents to compose multiple MCP tools (documentation search, fact-checking, prompt templates, problem analysis) into coordinated workflows for complex coding tasks. Implements tool chaining through MCP's function-calling interface, allowing agents to invoke tools sequentially or in parallel, pass results between tools, and maintain state across steps. Supports conditional branching based on tool results and error handling for failed tool invocations.