kubefwd

Automatically discovers and forwards multiple Kubernetes services from specified namespaces to the local machine, assigning each service a unique loopback IP address (127.x.x.x range) to eliminate port conflicts. Uses kubectl port-forward under the hood with intelligent IP allocation logic that maps services to distinct loopback addresses, enabling simultaneous access to services that would normally compete for the same local ports. This architecture allows developers to access multiple services without manual port remapping or conflict resolution.

Accepts natural language queries to discover, filter, and retrieve information about forwarded Kubernetes services without requiring kubectl syntax knowledge. Parses user intent (e.g., 'show me all database services' or 'which services are using port 5432') and translates it into service queries using label selectors, namespace filters, and port matching logic. Integrates with an LLM backend (via MCP) to interpret ambiguous queries and map them to concrete Kubernetes resource attributes.

Continuously monitors the health and connectivity status of active port forwards, detecting broken connections, network issues, and service unavailability. Implements health checks that periodically attempt connections to forwarded service endpoints and reports latency, error rates, and connection state. Provides diagnostic output including error logs, network trace information, and suggestions for remediation (e.g., 'service pod is not running' or 'network policy blocking traffic').

Captures and aggregates network traffic metrics for forwarded services, including request counts, latency distributions, error rates, and bandwidth usage. Collects metrics from kubectl port-forward process statistics and optional service instrumentation (Prometheus, OpenTelemetry), then presents them through natural language summaries or structured data exports. Enables developers to understand service usage patterns and performance characteristics without setting up external monitoring infrastructure.

Provides a conversational interface for managing port forwards using natural language commands interpreted by an LLM backend via MCP. Accepts commands like 'forward all services in the payments namespace' or 'stop forwarding the database service' and translates them into kubefwd operations. Maintains conversation context to handle follow-up commands and clarifications, enabling multi-turn interactions for complex port-forward management tasks.

Exposes kubefwd capabilities as MCP (Model Context Protocol) tools that can be composed with other AI agents and LLM applications. Implements MCP tool schemas for port-forward operations, service queries, and diagnostics, enabling external agents to invoke kubefwd functionality as part of larger AI workflows. Supports bidirectional communication with MCP servers to receive commands and report results in a standardized format.

Stores port-forward configurations (service mappings, loopback IP assignments, health check settings) persistently and automatically restores them on restart. Implements a configuration file format (YAML or JSON) that captures the desired state of all forwarded services, enabling developers to version-control their port-forward setup and share it across team members. Includes recovery logic that detects stale port-forward processes and re-establishes connections on restart.

Analyzes port-forward failures and connection errors using LLM-powered reasoning to identify root causes and suggest remediation steps. Correlates error messages with Kubernetes state (pod status, resource availability, network policies) and provides actionable debugging guidance. Integrates with kubectl to fetch cluster state and service logs, enabling comprehensive diagnosis without requiring developers to manually investigate.