# hound

AI-powered security analysis system with knowledge graph backend

- **Kind:** MCP server
- **Owner:** scabench-org
- **Source:** https://github.com/scabench-org/hound
- **Version:** 2.0.0
- **Runtime:** python
- **License:** Other
- **Pinned commit:** `83ce761952b00d9090ac0911b5e618bf3753087f`
- **Page:** https://forefy.com/mcps/b75a3611-f35c-5555-9585-28b4c67ef972
- **API (JSON):** https://forefy.com/api/asr/mcps/b75a3611-f35c-5555-9585-28b4c67ef972

## Install

Ask your agent: `Install the hound MCP server securely from: https://forefy.com/mcps/b75a3611-f35c-5555-9585-28b4c67ef972 and register it with my agent, prompting me for OPENAI_API_KEY, HOUND_GEMINI_MAX_OUTPUT_TOKENS`

## Tools (17)

- `set_emotion`: Deprecated: use set_avatar_state instead. Sets legacy UI emotion label.
- `set_avatar_state`: Set Inuzumi avatar state (not spoken).
- `get_hound_status`: Return a short status of Hound audits.
- `human_status`: Human-friendly audit status summary (no raw counts).
- `enqueue_steering`: Queue a steering note for the strategist.
- `list_plan`: List last planned investigations with done status.
- `get_current_activity`: What is being investigated right now.
- `list_hypotheses`: Top hypotheses (filterable)
- `get_system_overview`: SystemOverview graph stats and top nodes by degree.
- `search_graph_nodes`: Search nodes by label/type substring in SystemOverview.
- `get_node_details`: Get compact details + incident edges for a node.
- `get_file_snippet`: Fetch a small code snippet for a relpath or card_id (uses card store).
- `get_hypothesis_details`: Fetch full hypothesis details by id (node_refs, files, reasoning).
- `get_top_hypothesis`: Fetch the highest-confidence hypothesis with enriched details.
- `list_nodes`: List nodes from the SystemOverview graph.
- `list_files`: List known file relpaths (optionally filter by substring).
- `search_repo`: Search repository files for a query and return top file hits with snippets.

## Environment

- `HOUND_DEBUG_DIR`
- `OPENAI_BASE_URL`
- `HOUND_REGISTRY_DIR`
- `OPENAI_API_KEY` (secret)
- `PYTHON`
- `HOST`
- `PORT`
- `FLASK_DEBUG`
- `JOBLIB_MULTIPROCESSING`
- `HOUND_LLM_VERBOSE`
- `HOUND_GEMINI_REQUIRE_NEW`
- `HOUND_GEMINI_NO_LEGACY`
- `GOOGLE_USE_VERTEX_AI`
- `VERTEX_PROJECT_ID`
- `GOOGLE_CLOUD_PROJECT`
- `GCP_PROJECT`
- `VERTEX_LOCATION`
- `GOOGLE_CLOUD_REGION`
- `GOOGLE_CLOUD_ZONE`
- `HOUND_GEMINI_MAX_OUTPUT_TOKENS` (secret)
- `LITELLM_API_BASE`
- `HOUND_OPENAI_USE_RESPONSES`
- `HOUND_CONFIG`

