OpsCat: lightweight software catalog with AI agent integration launches

OpsCat is a single-binary software catalog platform released as an alternative to legacy developer portals like Backstage. Written in Go, it includes an embedded React/Vite web UI and runs in under 30MB RAM with no external runtime dependencies—a deliberate contrast to Backstage, which requires weeks of TypeScript boilerplate, complex yarn monorepos, and over 1GB of memory.

The tool operates through three main capabilities. First, it auto-discovers microservices and infrastructure by crawling codebases to infer tech stacks (Go, Node, Python, Rust, Java), containers (Dockerfile), infrastructure (Kubernetes, Helm, Terraform), and APIs (OpenAPI, GraphQL, Protocol Buffers). Second, it visualizes service relationships in an interactive 2D dependency graph using Vis.js, showing software relationships, systems, and resource dependencies in real time. Third, it automatically calculates compliance scorecards for each service on a 0-100% scale based on documentation, ownership, containerization, and API specifications.

Getting started requires only three steps: build from source using Go, run a scan command to auto-discover services and save to SQLite, then launch the web portal with a single serve command. The CLI includes commands for scanning workspaces, exporting catalogs in JSON/YAML/table format, and querying entities. Teams can optionally create a catalog.yaml manifest in any repository root to override or enrich auto-discovered metadata—specifying service name, owner, system, domain, tags, links to dashboards and repositories, and explicit dependencies.

A core feature is its native Model Context Protocol (MCP) server, which exposes four tools for AI coding agents: catalog_list_services (list microservices with filtering), catalog_get_service (retrieve detailed metadata), catalog_scan_workspace (trigger discovery dynamically), and catalog_get_dependencies (query upstream and downstream relationships). AI agents like Claude, Cursor, Antigravity, and Devin can be configured to connect to OpsCat via stdio or HTTP, allowing them to understand platform topology when generating code. OpsCat is distributed under the Apache 2.0 License and available on GitHub.

OpsCat addresses a specific pain point in platform engineering: developer portals have become unnecessarily heavy and time-consuming to set up. Backstage, the reference implementation in this space, requires extensive TypeScript boilerplate, monorepo management, and high memory footprint—friction that delays teams from adopting service catalog practices. By reimagining the problem in Go with embedded UI and zero external dependencies, OpsCat reduces setup friction to near-zero.

The release also reflects a shift in how AI agents interact with engineering systems. Rather than treating the catalog as a read-only view for humans, OpsCat exposes service topology, ownership, and health metrics through the Model Context Protocol—a specification designed to give AI coding agents structured context. This allows tools like Claude and Cursor to understand service dependencies when generating code, reducing context switching and enabling agents to make architecture-aware decisions.

The zero-configuration auto-discovery is significant: OpsCat crawls a codebase and infers services, containers, infrastructure, and APIs without manual setup. Optional catalog.yaml files allow teams to enrich metadata without requiring the setup burden of legacy platforms. This design choice lowers the barrier to adoption for teams of any size.

No comments yet. Be the first to share your thoughts!