Skip to content

Quick Start

This path installs Sentinel DV, indexes the checked-in demo artifacts, connects an MCP client, and verifies the server with real tool calls.

Prerequisites

  • Python 3.10 or newer
  • uv for the shortest install path
  • An MCP client such as Codex, Claude Code, or GitHub Copilot CLI

The commands below run v2.3.1 through uvx. For a persistent environment, use python3 -m pip install "sentinel-dv>=2.3.1".

1. Create a demo configuration

Clone the repository so the example artifacts and verification scripts are available:

git clone https://github.com/kiranreddi/sentinel-dv.git
cd sentinel-dv
cp demo/config.example.yaml demo/config.yaml

config.yaml is resolved relative to its own directory. Review artifact_roots and index.path before using the same pattern with production artifacts.

2. Index artifacts

uvx --from sentinel-dv@2.3.1 \
  sentinel-dv-index --config "$PWD/demo/config.yaml" --index-all

Indexing is a required, separate step. The MCP server reads the DuckDB index; it does not scan raw artifacts on each tool call.

3. Connect an MCP client

codex mcp add sentinel-dv \
  --env SENTINEL_DV_CONFIG="$PWD/demo/config.yaml" \
  -- uvx --from sentinel-dv@2.3.1 sentinel-dv-server

claude mcp add \
  --env SENTINEL_DV_CONFIG="$PWD/demo/config.yaml" \
  --transport stdio --scope local sentinel-dv \
  -- uvx --from sentinel-dv@2.3.1 sentinel-dv-server

copilot mcp add sentinel-dv \
  --env SENTINEL_DV_CONFIG="$PWD/demo/config.yaml" \
  -- uvx --from sentinel-dv@2.3.1 sentinel-dv-server

Use an absolute path for SENTINEL_DV_CONFIG. See Agent setup for configuration files, project scope, skills, and host-specific verification.

4. Verify the connection

In the client, confirm that the sentinel-dv server is connected and that runs.list is available:

Client Check
Codex /mcp
Claude Code /mcp or claude mcp list
GitHub Copilot CLI /mcp list or copilot mcp list

Then ask:

List the indexed runs with runs.list. Do not summarize beyond the returned data.

An empty result is not a passing regression. It means the selected index has no matching records or needs to be rebuilt.

5. Use a workflow skill

The repository includes three skills under .agents/skills, .claude/skills, and .github/skills.

Use the Sentinel DV regression triage skill to assess the latest failing run.
Preserve IDs, disclose missing health components, and prioritize failure clusters.

Use the Sentinel DV failure debugging skill to explain why test_counter_sim failed.
Separate observations, inference, and missing evidence.

Use the Sentinel DV coverage closure skill for the latest xcelium run.
Prioritize AXI4 functional gaps and check assertion vacuity.

The skills guide the tool sequence and reporting rules. The MCP server remains the source of live, typed verification data.

Repository verification

From a development checkout:

.venv/bin/python scripts/verify_all_mcp_tools.py
.venv/bin/python scripts/verify_skill_workflows.py

The first command invokes all 28 MCP tools. The second executes the three published skill workflows against 52 checked-in demo artifacts.

Next