Keep AI Context Fresh With Every Build
AI coding agents are only as useful as the context they can access. In large codebases, that context changes every time services, APIs, fields, consumers, or dependencies change. HoundDog.ai integrates directly into your CI workflows to continuously regenerate this context from source code, and the MCP server hands it to your agents on demand.
Add the MCP server to Claude Code
The server runs locally over stdio and reads the scan state store on your machine. No API key. No prior scan required. It works fully offline.
# Claude Code
claude mcp add hounddog -- hounddog mcp
One deterministic chain, recomputed as your code changes
Our fast, deterministic analysis engine runs on standard CPU infrastructure with minimal impact on CI latency. Instead of asking AI agents to repeatedly search repositories and reconstruct the same relationships, HoundDog.ai computes those relationships once and refreshes them as your code changes.
The resulting evidence can power your service catalog, API and gRPC and Thrift discovery, microservices documentation, change blast radius analysis, and organization wide context for AI coding agents.
Through MCP, authorized agents such as Claude Code can query this continuously updated context whenever they need it.
Static analysis discovers. AI reasons. MCP delivers.
Use AI where reasoning adds value. Let CI handle the deterministic discovery. The MCP server is the delivery layer in between: it serves deterministic, precomputed context instead of letting each agent rediscover it by grepping.
API catalog
Query gRPC and Thrift services, their methods, server-side implementations, and every client call site across your repositories.
# in your agent list_api_services(protocol="grpc") get_api_service(full_name="hipstershop.CartService")
Dataflows
Query sensitive data paths to their sinks, with severity, AI review verdicts, CWE and OWASP references, and remediation guidance attached.
# in your agent list_dataflows(min_severity="high") get_dataflow(hash="<dataflow-hash>")
Works with every MCP-compatible agent
Claude Code, Cursor, Codex, GitHub Copilot, JetBrains AI, Windsurf, OpenCode, and anything else that speaks the Model Context Protocol. The transport is stdio, so configuration is a command and its arguments.
Claude Code
Register the server once and it is available in every session in that project.
claude mcp add hounddog -- hounddog mcp
Any other MCP client
Point the client at the hounddog mcp stdio server.
{
"command": "hounddog",
"args": ["mcp"]
}
The tools your agent gets
Six typed tools, all reading from the local scan state store. Full parameter reference in the MCP documentation.
| Tool | Key parameters | What it returns |
|---|---|---|
get_scan_status | — | Store state, last scan metadata, and scanned-file counts. |
run_scan | — | Produces or refreshes stored results using differential scanning. |
list_api_services | protocol, name_contains | Services with method, server, and client counts plus severity data. |
get_api_service | full_name, protocol, client_offset | Detailed service info: methods, implementations, and dataflows. |
list_dataflows | min_severity, data_sink, file_path, limit, offset | Open dataflows sorted by effective severity. |
get_dataflow | hash | Full dataflow detail with code, CWE and OWASP references, and remediation. |
- Your code stays put. The server runs locally over stdio and reads the local scan state store. Scans triggered through MCP never upload results to the Cloud Platform.
- Large repositories: a first scan can exceed a client's tool timeout. Run
hounddog scanonce from the terminal, then let the agent work against the stored results. - Scan rules load at startup. Restart the server to pick up rule updates.
- Not a git repository? Use
--no-gitmode.
Compute the facts once. Keep them fresh.
Let every authorized AI agent reason over context that is already current, instead of paying in tokens, latency, and compute to rediscover it on every prompt.