Every gRPC service and protobuf message, mapped for you and your AI agents.

ProtoMap is a context engine and model harness specialized for gRPC and Protobuf (more protocols coming soon). It scans your source code and provides organization-wide, centralized context to engineers and coding agents. Plug in the ProtoMap MCP and benefit immediately. Benchmarks show 13× fewer tokens, 7.7× faster answers, 6.4× lower cost, and increased accuracy. The bigger the codebase, the bigger the gain.

Book a Demo Start Free

Paste into your coding agent

Set up ProtoMap and show me what it can do. ProtoMap maps gRPC services and protobuf messages across repositories, and gives coding agents that graph over MCP. Docs: https://docs.protomap.ai/quickstart/

Tell me what you are about to do before each step that installs something.

1. Check that this computer has git, Docker with Compose 2.23.1 or newer, at least 4 GB of RAM, and 20 GB of disk space. If anything is missing, tell me and stop.

2. Install ProtoMap Server, in a folder outside this repository:

   git clone https://github.com/hounddogai/protomap.git
   cd protomap
   ./self-hosted/scripts/install.sh

   The installer asks questions. Choose Trial, which runs PostgreSQL locally and logs the CLI in, and keep the default port.

3. Show me the address and the one-time setup key the installer prints. Ask me to open the address, enter the key, and create my organization and owner account. Wait until I say that's done.

4. Scan this repository, from a new shell so the CLI is on the PATH:

   protomap scan .

   If it has no gRPC services in a supported language, tell me, and offer to scan a public example instead:

   git clone https://github.com/GoogleCloudPlatform/microservices-demo.git
   protomap scan microservices-demo

5. From the scanned repository's folder, add ProtoMap to yourself as an MCP server:

   Claude Code: claude mcp add protomap -- protomap mcp serve
   Codex: codex mcp add protomap -- protomap mcp serve
   Antigravity: add "protomap": { "command": "protomap", "args": ["mcp", "serve"] } under mcpServers in ~/.gemini/config/mcp_config.json
   Others: https://docs.protomap.ai/coding-agents/

   Tell me if I need to restart you for it to take effect.

6. Tell me what the scan found, then use ProtoMap's tools to answer one question about my services, such as which services call a method and where. Suggest two more questions I could ask.

HoundDog.ai, the creator of ProtoMap, is trusted by

  • Replit
  • Labcorp
  • Wintrust
  • Senseonics
  • Juvare

Why ProtoMap

Fast, cheap, and accurate at scale

Stop burning tokens spinning up subagents and rediscovering the same code. Instruction and memory files go stale quickly and are challenging to maintain across large codebases. ProtoMap derives its context deterministically from the code itself, with no additional files to write or maintain. The analysis runs efficiently on a modern laptop, with no GPU required. Future-proof against the inevitable explosion of code.

Example Prompt

“We want to remove the gRPC method PaymentService/CreatePayment. Find every place in every repo that calls it, including indirect calls.”

Claude Working… Answered
  1. Bash grep -rn "CreatePayment|create_payment" across all 9 repos
  2. Bash Count the files in each repository
  3. Bash Count the files mentioning create_payment, per repo
  4. Bash grep -rIc -i "create_?payment" | sort
  5. Thinking Direct mentions are only a start; tracing helpers, registries, and string-built names repo by repo
  6. Bash Python: grep "getattr|registry[" in src
  7. Bash Python: show every getattr call site
  8. Bash Python: grep "unary_unary|acme.payments.v1."
  9. Thinking Python done: 3 real callers. Moving on to Go
  10. Bash Go: grep "CreatePayment" outside generated code
  11. Bash Go: grep ".Invoke(|MethodByName|reflect."
  12. Bash grep every repo for "Create" and "Payment" string pieces
  13. Grep Method names built from strings, across all repos
  14. Bash pkill the slow background grep
  15. Thinking Go has 4 callers. Sweeping the other 7 repos with a broader pattern
  16. Grep "create.?payment", any case, all repos
  17. Bash C#: grep hand-built "new Method<" descriptors
  18. Thinking C# has 4 callers, one through a hand-built descriptor. JS and TS next
  19. Bash JS: grep "paymentClient|PaymentService"
  20. Thinking JS has 2 callers, one through a string key. TypeScript next
  21. Bash TS: grep "Client[|makeUnaryRequest|as keyof"
  22. Thinking TypeScript has 1 caller. Now Swift and Rust
  23. Bash Swift: grep "MethodDescriptor(|methodName:"
  24. Bash Rust: grep "PathAndQuery|format!("/{}/{}")"
  25. Thinking Rust has 2 callers. Kotlin and Java match nothing by name, so looking for indirect forms
  26. Bash Java, Kotlin: grep "generateFullMethodName|getDeclaredMethod"
  27. Thinking Java builds method names from split strings, like "Cre" + "atePayment". Searching for fragments
  28. Grep Every split of "CreatePayment" into two strings
  29. Grep "findMethodByName|Class.forName|ServerReflection"
  30. Thinking Java has no callers. Confirming the handler around each call site
  31. Bash awk: the enclosing function of each call site
  32. Bash grep compiled files (.pyc, .class, .dll) for createpayment
  33. Bash awk: the enclosing C# handlers
  34. Bash grep the callers of the Swift and Rust helpers
  35. Answer 18 call sites in 8 of the 9 repos, with file:line
Time
8m 13s
Tokens
7.0M
Cost
$4.46
Claude + ProtoMap Working… Answered
  1. ToolSearch Load ProtoMap’s list_call_sites tool
  2. protomap list_call_sites rpc: acme.payments.v1.PaymentService/CreatePayment
  3. protomap list_call_sites unresolved: true
  4. Grep "CreatePayment|create_payment", count per file
  5. Read checkout/…/PaymentSyncGrpcService.kt
  6. Grep "workerForSettlePayments0" in checkout
  7. Read checkout/…/PaymentSyncGrpcService.kt
  8. Answer 18 call sites in 8 of the 9 repos, with file:line
Time
27s
Tokens
130K
Cost
$0.20
From our benchmarks on a synthetic codebase (5M LOC), Opus 5.5 at high effort: totals averaged over two runs, steps from one of them, abridged, paraphrased, and sped up.

Private by Design

Runs on-prem with Docker and PostgreSQL. Bring your own model for complete isolation.

Enterprise Ready

Scans millions of lines of code in minutes. Supports both monorepos and multi‑repos.

Security First

HoundDog.ai is SOC 2 compliant, pentested regularly, and a member of Anthropic’s Defense Access and OpenAI’s Daybreak Blue programs.

How it works

From source code to answers, in four steps.

Scan

Run protomap scan to analyze your gRPC and Protobuf code. A million lines of code in seconds.

Supported Languages

  • Go
  • Java
  • Kotlin
  • C#
  • Python
  • TypeScript
  • JavaScript
  • Rust
  • Swift
  • Monorepos
  • Multi-repos

Graph

Entities and their connections form a graph, updated on every merge to main and in real time for local code changes.

Example Entities Tracked

  • Repos
  • Services
  • Methods
  • Fields
  • Messages

Review

Bring your own AI model to prune false positives.

Supported AI Providers

  • Anthropic
  • OpenAI
  • Grok
  • TypeSafe.ai
  • Gemini
  • AWS Bedrock
  • Microsoft Foundry

Query

Agents query the graph over the ProtoMap MCP.

Supported Coding Agents

  • Claude Code
  • Codex
  • Antigravity
  • VS Code
  • OpenCode
  • + any MCP client

Humans query it through the model harness.

Example Queries

  • blast radius
  • field lineage
  • call graph
  • dependencies
  • service hotspots
  • dead gRPC methods
  • code owners

Backed by benchmarks

A code context graph that actually works, measured on both synthetic and real codebases with millions of lines of code.

faster per prompt
7.7×
cheaper per prompt
6.4×
fewer tokens
13×
Measure
Opus 5.5, high effort
Prompt Claude Code time Claude Code + ProtoMap Faster Claude Code cost Claude Code + ProtoMap Cheaper Claude Code tokens Claude Code + ProtoMap
Services in a repo 1m 12s 39s 1.8× $0.58 $0.31 1.9× 411K 109K
Most-connected services 23m 56s 12s 121× $9.86 $0.15 64× 15.8M 71K
Field blast radius 6m 27s 21s 18× $9.00 $0.22 41× 16.8M 107K
Changing a request field 3m 24s 1m 09s 3.0× $1.69 $0.65 2.6× 1.9M 693K
Response field source 25s 18s 1.4× $0.24 $0.21 1.2× 172K 161K
Two-hop field lineage 49s 24s 2.0× $0.47 $0.20 2.4× 378K 121K
Removing a method 8m 13s 27s 18× $4.46 $0.20 23× 7.0M 130K
Service outage impact 7m 18s 16s 28× $1.97 $0.13 15× 3.4M 68K
Shortest call chain 5m 09s 8s 39× $1.55 $0.076 20× 2.0M 62K
Callers outside handlers 4m 59s 28s 11× $8.70 $0.20 44× 17.1M 100K
Service API docs 1m 54s 22s 5.3× $1.06 $0.23 4.6× 1.3M 127K
Dead gRPC methods 2m 49s 2m 41s 1.1× $0.97 $0.51 1.9× 1.1M 502K
What a handler calls 2m 32s 36s 4.2× $0.34 $0.39 0.9× 261K 372K
Cross-domain calls 2m 54s 20s 8.7× $0.92 $0.19 4.8× 1.0M 86K
Proto change impact 2m 11s 24s 5.4× $1.51 $0.31 4.9× 1.6M 139K
All prompts 96m 55s 10m 42s $50.69 $5.13 80.5M 3.4M

From our benchmarks on a synthetic codebase (5M LOC). Opus 5.5 at high effort, on a log scale. The chart shows 15 key prompts of 20; the figures above cover all 20. See every prompt in the full benchmarks. Numbers are subject to change with testing on more codebases and newer models.

Beyond gRPC, coming soon

ProtoMap is expanding to more protocols, with end-to-end data lineage across the entire stack.

REST APIs

OpenAPI, endpoints, and fields

GraphQL

Schemas, resolvers, and fields

Apache Thrift

IDL files, services, and fields

Full Data Lineage

Source to sink, across your stack

See it in action

HoundDog.ai ProtoMap live demo at MozFest 2026 10:53 Live demo · MozFest 2026

HoundDog.ai ProtoMap Demo at MozFest

Live demo of HoundDog.ai’s ProtoMap at MozFest’s Open Source House: callers and callees, and the blast radius of changing a field name.

 HoundDog.ai Benchmarks · Side by side ProtoMap × Claude Code 7× faster · 36× cheaper Claude Code alone6:01 · $14.44 Claude Code + ProtoMap0:50 · $0.40 1:09 Benchmarks · Video

ProtoMap × Claude Code, side by side

The same prompt in two live sessions: ProtoMap’s answers in 50 seconds for $0.40, the baseline’s in 6 minutes for $14.44.

Try it on your code.

Book a Demo Start Free

Paste into your coding agent

Set up ProtoMap and show me what it can do. ProtoMap maps gRPC services and protobuf messages across repositories, and gives coding agents that graph over MCP. Docs: https://docs.protomap.ai/quickstart/

Tell me what you are about to do before each step that installs something.

1. Check that this computer has git, Docker with Compose 2.23.1 or newer, at least 4 GB of RAM, and 20 GB of disk space. If anything is missing, tell me and stop.

2. Install ProtoMap Server, in a folder outside this repository:

   git clone https://github.com/hounddogai/protomap.git
   cd protomap
   ./self-hosted/scripts/install.sh

   The installer asks questions. Choose Trial, which runs PostgreSQL locally and logs the CLI in, and keep the default port.

3. Show me the address and the one-time setup key the installer prints. Ask me to open the address, enter the key, and create my organization and owner account. Wait until I say that's done.

4. Scan this repository, from a new shell so the CLI is on the PATH:

   protomap scan .

   If it has no gRPC services in a supported language, tell me, and offer to scan a public example instead:

   git clone https://github.com/GoogleCloudPlatform/microservices-demo.git
   protomap scan microservices-demo

5. From the scanned repository's folder, add ProtoMap to yourself as an MCP server:

   Claude Code: claude mcp add protomap -- protomap mcp serve
   Codex: codex mcp add protomap -- protomap mcp serve
   Antigravity: add "protomap": { "command": "protomap", "args": ["mcp", "serve"] } under mcpServers in ~/.gemini/config/mcp_config.json
   Others: https://docs.protomap.ai/coding-agents/

   Tell me if I need to restart you for it to take effect.

6. Tell me what the scan found, then use ProtoMap's tools to answer one question about my services, such as which services call a method and where. Suggest two more questions I could ask.

In the self-hosted version, no source code leaves your environment. If you use our cloud version, only the minimal code context around tracked entities is transmitted for analysis, never full repositories. It is deleted after analysis and never used for training.