Installation¶
Sentinel DV v2.3.1 is distributed through PyPI and the MCP Registry as io.github.kiranreddi/sentinel-dv.
Choose an install¶
uvx runs an isolated package without modifying the active Python environment:
Python 3.10 or newer is required.
Configure¶
Copy the production template:
At minimum, set:
artifact_roots:
- /absolute/path/to/regression/artifacts
index:
type: duckdb
path: ./sentinel_dv.db
adapters:
uvm: true
cocotb: true
assertions: true
coverage: true
waveform_summary: true
Relative paths are resolved from the directory containing the config file. See Configuration for security limits, redaction, submit templates, and adapter settings.
Build the index¶
Run indexing after artifacts change. The server queries the index and does not watch artifact roots.
Start the stdio server¶
A stdio MCP server normally appears silent when started directly because stdout is reserved for JSON-RPC. Use your MCP client's status view to test the handshake.
You may set the config once instead:
If neither --config nor SENTINEL_DV_CONFIG is set, Sentinel DV checks for config.yaml or config.yml in the server's working directory. It never falls back to demo artifacts in production.
Connect an agent¶
Continue to Agent setup for Codex, Claude Code, GitHub Copilot CLI, and generic MCP JSON examples.
Verify a development checkout¶
Expected final lines:
These checks index the checked-in demo corpus; they do not require commercial simulator licenses.
Upgrade¶
Pin the package version in MCP configuration so every host runs the same server:
After changing versions, rebuild the index and re-run the connection check. Review the changelog before adopting schema or tool changes.