ProtoMap
A context engine and model harness that maps gRPC services and Protobuf messages across repositories and serves that context to engineers and coding agents over MCP.
At a Glance
About ProtoMap
ProtoMap is made by HoundDog.ai. It scans source code to build an organization-wide map of gRPC services, Protobuf messages, methods and fields, and exposes that map to engineers and AI coding agents. The vendor lists a free self-hosted option on GitHub, and cloud or on-premises deployment for organizations.
What It Is
ProtoMap is a context engine and model harness specialized for gRPC and Protobuf. Running protomap scan analyzes a monorepo or multiple repositories and builds a graph of repos, services, methods, messages and fields. The vendor states that the context is derived deterministically from the code itself, needs no AI tokens for analysis, and runs on a modern laptop without a GPU.
How the Workflow Runs
The documented flow has four steps: scan, graph, review, query. Supported scan languages are Go, Java, Kotlin, C#, Python, TypeScript, JavaScript, Rust and Swift. The graph is updated on every merge to main and in real time for local code changes. An optional review step lets users bring their own AI model (Anthropic, OpenAI, Grok, TypeSafe.ai, Gemini, AWS Bedrock or Microsoft Foundry) to prune false positives. Coding agents then query the graph through the ProtoMap MCP server, and humans query it in plain English through the model harness.
Questions It Answers
Example queries include blast radius of changing a service or field, method callers and call sites, field lineage, dependencies, service hotspots, dead gRPC methods and code owners. Answers include file and line locations for call sites.
Benchmarks and Deployment
HoundDog.ai reports benchmarks on a synthetic 5M-LOC codebase of 9 gRPC repositories using Claude Code with Opus 5.5: 13x fewer tokens, 7.7x faster answers and 6.4x lower cost, with results varying by prompt (some prompts showed little or negative gain). The vendor notes numbers may change with more codebases and newer models. The self-hosted server runs with Docker and PostgreSQL, and the vendor states no source code leaves the environment in that mode. The cloud version transmits only minimal code context around tracked entities. The vendor lists REST/OpenAPI, GraphQL, Apache Thrift and full data lineage as upcoming.
Community Discussions
Be the first to start a conversation about ProtoMap
Share your experience with ProtoMap, ask questions, or help others learn from your insights.
Pricing
Local Scanner
Free local ProtoMap scanner that runs on your machine and plugs into any MCP-compatible AI agent.
- Scans whatever code is available on your machine
- Provides local API graphs for documentation
- Works with any MCP-compatible AI agent
- Available as MCP Servers
- Supports gRPC (GraphQL and REST coming soon)
Enterprise
Centralized context for the entire organization, continuously updated across all code repos; cloud or on-prem.
- Cloud (fully managed SaaS) or On-Premises
- SOC-2 compliant, with SSO, RBAC, and audit logs
- Direct source code integrations with GitHub, Bitbucket, and GitLab
- Auto-scans selected repos, no local checkout needed
- Runs scans in CI and on every pull request
- Keeps API graphs continuously updated org-wide
Capabilities
Key Features
- Scans gRPC and Protobuf code across monorepos and multi-repos
- Graph of repos, services, methods, messages and fields updated on every merge and for local changes
- MCP server for coding agents
- Blast radius analysis for services and fields
- Method callers and call sites with file locations
- Field-level lineage and dependencies
- Dead gRPC method detection
- Code owner and dependent lookup
- Optional AI review to prune false positives with a bring-your-own model
- Self-hosted deployment with Docker and PostgreSQL
- Plain-English querying through the model harness
- CI refresh of context on every commit
