CircleCI

Implements the Model Context Protocol (MCP) v1.8.0 specification to register CircleCI tools with MCP-enabled clients, enabling dynamic tool discovery and invocation through a standardized schema-based interface. The server maintains a tool registry where each tool is registered with name, description, input schema, and handler function, allowing LLM clients to discover available CircleCI operations and invoke them with proper type validation.

Retrieves detailed failure logs from CircleCI builds via the CircleCI v2 API and extracts structured context including error messages, stack traces, and failure timestamps. The system parses raw build logs to identify failure patterns and provides them to LLM agents for root-cause analysis and remediation suggestions, supporting both direct CircleCI URLs and local git repository context for project detection.

Analyzes CircleCI test results across multiple pipeline executions to identify flaky tests (tests that fail intermittently) using statistical patterns and historical data. The system queries the CircleCI v2 API to retrieve test results, correlates failures across runs, and provides structured data about test reliability metrics, failure frequency, and affected test suites to enable targeted remediation.

Queries CircleCI v2 API to retrieve the current status of pipelines for a given project, including workflow status, job statuses, and pipeline metadata. The system fetches the latest pipeline execution or specific pipeline by ID, providing real-time visibility into CI/CD pipeline state and enabling agents to monitor build progress, detect stuck pipelines, and trigger downstream actions based on pipeline status.

Retrieves detailed test results from CircleCI jobs via the v2 API, including test names, durations, status (passed/failed), and failure messages. The system parses test result data into structured format and correlates it with job metadata, enabling agents to analyze test performance, identify slow tests, and extract failure details for debugging or test remediation workflows.

Implements a project detection system that resolves CircleCI project identifiers from either explicit CircleCI URLs or local git repository context (git remote origin URL). The system parses git remote URLs to extract organization and project names, enabling tools to work without requiring users to explicitly provide CircleCI project slugs, reducing configuration friction and supporting context-aware operations.

Provides a unified API client abstraction that supports both CircleCI v1.1 (legacy) and v2 (current) APIs, with automatic fallback to private API endpoints when needed for operations not available in public APIs. The system handles authentication, request formatting, and response parsing for multiple API versions, enabling tools to access CircleCI functionality regardless of API version availability and providing graceful degradation when public APIs lack required features.

Provides pre-built prompt templates that guide LLM agents through structured interaction patterns for common CircleCI tasks (e.g., debugging build failures, analyzing test results). These templates define expected input/output formats, reasoning steps, and context requirements, enabling more reliable and consistent agent behavior when performing CircleCI operations through natural language.

Defines comprehensive JSON schemas and TypeScript type definitions for all tool inputs and CircleCI API responses, enabling input validation, IDE autocomplete, and type-safe tool invocation. The system validates tool inputs against schemas before API calls, catching configuration errors early and providing clear error messages, while type definitions enable developers to build type-safe integrations without manual schema parsing.

Provides utility functions for parsing, transforming, and formatting CircleCI API responses into human-readable and agent-friendly formats. The system handles common data transformations (e.g., converting timestamps to readable dates, extracting error messages from nested responses, formatting test results into tables), reducing the burden on tools and agents to manually parse API responses.