pal-mcp-server

Implements a ModelProviderRegistry pattern that abstracts 7+ distinct AI providers (Gemini, OpenAI, Azure, Grok, OpenRouter, DIAL, Ollama, custom endpoints) behind a single interface. Each provider implements a common contract with native API bindings, enabling seamless switching and fallback without client-side provider logic. The abstraction handles provider-specific authentication, request formatting, response normalization, and error handling through a registry-based dependency injection pattern.

Maintains conversation continuity across MCP context resets using a continuation-based reconstruction pattern stored in _conversation_memory. When context is lost (e.g., token limits exceeded), the system reconstructs prior conversation state by replaying message history through reconstruct_thread_context() without requiring persistent external storage. This enables multi-turn workflows in stateless MCP environments where clients cannot maintain session state between requests.

Provides a planner tool that decomposes complex development tasks into actionable steps with dependencies and resource requirements. The tool analyzes task descriptions, identifies prerequisites, estimates effort, and creates execution plans that can be executed sequentially or in parallel. It integrates with other tools (refactor, test generation, security audit) to create comprehensive workflows.

Integrates web search capabilities into the MCP server, enabling tools to fetch current information, documentation, and examples from the internet. When analyzing code or generating solutions, tools can search for relevant documentation, API references, security advisories, and best practices. Search results are incorporated into model context to provide up-to-date information beyond the model's training data.

Provides a tracer tool that captures detailed execution traces of code execution, including function calls, variable states, and control flow. The tool instruments code or integrates with debuggers to collect execution data, then presents it to AI models for analysis. This enables AI-assisted debugging where the model can inspect execution traces and identify root causes of bugs.

Provides a precommit tool that integrates with Git pre-commit hooks to run automated code quality checks before commits. The tool can execute code review, security audit, test generation, and other analysis tools on staged changes, blocking commits that fail quality gates. It provides fast feedback to developers and prevents low-quality code from entering the repository.

Provides a debug tool that helps diagnose and fix code issues through interactive analysis. The tool accepts error messages, stack traces, or problem descriptions, then uses AI reasoning to identify root causes and suggest fixes. It can integrate with execution traces and code context to provide targeted debugging assistance.

Provides an API lookup tool that searches and retrieves API documentation for libraries, frameworks, and services used in code. The tool can identify API calls in code, fetch relevant documentation, and provide context to models for code generation and analysis. It supports multiple documentation sources (official docs, OpenAPI specs, type definitions) and integrates with web search for current information.

Implements the Model Context Protocol as a stdio-based JSON-RPC server (Server('pal-server') in server.py) that exposes a registry of 10+ specialized tools for code analysis, debugging, and development workflows. The server handles tool discovery (listing available tools with schemas), parameter validation, and invocation routing through a unified TOOLS registry. Clients like Claude Code and Gemini CLI discover tools via MCP initialization and invoke them with structured parameters, receiving results back through the MCP protocol.

Implements an auto-mode strategy that selects the best available model based on task requirements, provider availability, and cost/performance tradeoffs. When a primary model is unavailable (rate-limited, API down, quota exceeded), the system automatically falls back to alternative providers without user intervention. The fallback logic considers model capabilities (vision, function calling, reasoning depth) and provider-specific constraints, routing requests intelligently across the provider registry.

Provides a consensus tool that invokes multiple AI models on the same code review task and synthesizes their outputs into a unified recommendation. Each model analyzes code independently, and the consensus engine identifies agreement patterns, flags disagreements, and produces a final review that incorporates diverse perspectives. This reduces false positives from single-model analysis and improves review quality by leveraging model diversity.

Implements a ThinkDeep tool that leverages extended thinking modes (where supported by models like Claude with extended thinking) to enable step-by-step reasoning for complex problems. The tool routes requests to models with reasoning capabilities, captures intermediate reasoning steps, and returns both the reasoning trace and final answer. This enables transparent, auditable AI decision-making for tasks like architecture design, debugging strategies, and security analysis.

Provides a test generation tool that analyzes code files and generates unit tests with full context awareness. The tool extracts function signatures, dependencies, and existing test patterns, then generates tests that match the codebase's testing conventions. It supports multiple testing frameworks (pytest, unittest, Jest, etc.) and can generate tests for specific functions or entire modules.

Provides a refactor tool that breaks down complex refactoring tasks into multiple steps, executing transformations incrementally while validating each step. The tool can handle large-scale refactorings (e.g., renaming across multiple files, extracting functions, modernizing syntax) by decomposing them into smaller, testable changes. It integrates with the planner tool to create refactoring strategies before execution.

Provides a security audit tool that analyzes code for common vulnerabilities, security anti-patterns, and compliance issues. The tool performs static analysis using AI models to identify issues like SQL injection risks, insecure cryptography, hardcoded secrets, and authentication flaws. It generates detailed reports with severity levels, affected code locations, and remediation recommendations.

Provides a documentation generation tool that analyzes code and produces comprehensive documentation including API docs, architecture guides, and usage examples. The tool extracts function signatures, docstrings, type hints, and code structure to generate documentation in multiple formats (Markdown, HTML, Sphinx). It can generate both reference documentation and narrative guides explaining system design and usage patterns.