ProtoMap

Product Introduction

1. Definition: ProtoMap is a deterministic gRPC and Protobuf context engine and model harness developed by HoundDog.ai. Operating as a specialized code intelligence tool, it maps every service, message, and field across multi-repository architectures, providing real-time, organization-wide structural awareness for both human engineers and AI coding agents via a Model Context Protocol (MCP) server.
2. Core Value Proposition: ProtoMap exists to eliminate the high latency, token consumption, and inaccuracy of traditional code search mechanisms—such as iterative grep commands or stale markdown documentation—when planning architecture changes in microservices. By generating a deterministic cross-repo dependency graph updated on every commit, ProtoMap enables AI coding agents like Claude Code to perform accurate impact analysis, service discovery, and field lineage tracking with significantly lower token costs and faster query responses.

Main Features

1. Deterministic Code Scanning & Graph Generation: ProtoMap analyzes source code across nine programming languages (Go, Java, Kotlin, C#, Python, TypeScript, JavaScript, Rust, and Swift) and supports both monorepos and multi-repo setups. Running via local CLI or CI/CD integrations, it parses Protobuf definitions (.proto) and gRPC implementations to build an exact graph of repositories, services, methods, messages, and fields without relying on probabilistic AI guesswork for the initial extraction.
2. AI Model Harness for False Positive Pruning: To handle complex or indirect call sites—such as reflection, dynamic method invocation, or string-built method names—ProtoMap incorporates an AI review layer. It allows developers to bring their own AI model providers (including Anthropic, OpenAI, Gemini, Grok, and AWS Bedrock) to validate potential call paths and prune false positives, balancing deterministic accuracy with semantic understanding.
3. Model Context Protocol (MCP) Server Integration: ProtoMap exposes its dependency graph directly to AI coding environments through an MCP server interface. Coding agents query the ProtoMap MCP using standardized tools (list_call_sites, service discovery, field blast radius queries) rather than repeatedly reading raw source files, drastically reducing redundant token usage and operational overhead during agentic coding workflows.

Problems Solved

1. Pain Point: Software engineers and AI agents working within large microservice architectures waste thousands of AI tokens and minutes of compute time executing recursive file searches, reading outdated documentation, or manually tracing gRPC method calls and Protobuf field dependencies across fragmented repositories.
2. Target Audience: Backend Engineers, Distributed Systems Architects, Platform Engineers, DevOps Teams, and organizations heavily utilizing gRPC/Protobuf service-to-service communication alongside AI coding assistants.
3. Use Cases: Determining the complete blast radius of modifying a database field or Protobuf message definition (e.g., changing Account.currency_code), identifying dead gRPC methods across enterprise services, tracing cross-domain API calls, and generating up-to-date service catalogs and API context for onboarding or refactoring tasks.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Pros: Delivers measurable performance gains (benchmarked at up to 13× fewer tokens, 7.7× faster answers, and 6.4× lower cost compared to raw Claude Code execution); updates deterministically on every commit; runs securely on-prem with Docker and PostgreSQL; supports air-gapped environments with bring-your-own-model configurations.
   • Cons: Currently specialized primarily for gRPC and Protobuf systems (though other protocols are on the roadmap); requires initial setup of the Docker-based self-hosted server and CLI indexing within the engineering organization's workflow.
2. Key Alternatives & Differentiation:
   • Traditional Code Search (grep/ripgrep): Requires manual aggregation, struggles with indirect/reflected method calls, and provides zero structural context for AI agents.
   • Static Documentation & Service Catalogs (e.g., Backstage): Frequently go stale because they depend on manual developer updates rather than being derived directly from source code on every commit.
   • General-Purpose Code Indexers / RAG Solutions: Rely heavily on vector embeddings and chunking, which often miss exact cross-repo API contracts and burn excessive tokens searching irrelevant code files. ProtoMap differentiates itself by providing a structured, deterministic graph specifically engineered for API dependencies and gRPC semantics.

Frequently Asked Questions (FAQ)

1. Question? How does ProtoMap reduce AI coding agent token consumption?
   • Answer: ProtoMap acts as an MCP server that provides AI coding agents with direct, structured answers about gRPC services, API contracts, and field dependencies. Instead of letting an agent recursively read files, execute broad grep commands, and load entire repositories into its context window, the agent queries the ProtoMap graph directly, cutting token usage by up to 13×.
2. Question? Which programming languages and protocols does ProtoMap support?
   • Answer: ProtoMap currently supports gRPC and Protobuf systems across nine languages: Go, Java, Kotlin, C#, Python, TypeScript, JavaScript, Rust, and Swift. Additional communication protocols and language parsers are continuously added to its roadmap.
3. Question? Is ProtoMap secure for enterprise codebases with strict compliance requirements?
   • Answer: Yes. ProtoMap is designed for enterprise readiness, offering on-premise deployment via Docker and PostgreSQL. It allows organizations to bring their own AI models for verification, ensuring that proprietary source code and dependency graphs remain completely isolated within the organization's infrastructure. HoundDog.ai is also SOC 2 compliant.
4. Question? How is ProtoMap integrated into existing CI/CD pipelines and developer workflows?
   • Answer: Engineers install the ProtoMap CLI (protomap scan .) locally or integrate it into CI pipelines to update the dependency graph on every merge to main. Developers then add ProtoMap as an MCP server to their local coding environments (such as Claude Code or VS Code) using simple configuration commands, allowing agents to query the graph instantly during development.