Skip to main content
Pidgeon’s public beta gives engineers three ways into the same local-first workflow: the .NET engine packages, the pidgeon command-line tool, and a thin Model Context Protocol adapter. Start with the CLI unless you are embedding the engine into an application or connecting an MCP client.
These are beta packages. Pin the exact versions below, review the known limits, and test the workflow before relying on it in a controlled environment. The desktop products are distributed separately from these community packages.

​
Choose an entry point

​
Run the CLI locally

Install the .NET tool:
Generate a deterministic synthetic HL7 v2 message, validate it, and preserve the output for review:
The seed makes the generated example repeatable. The validation result is evidence about that input and the supported rules, not a certification or a guarantee about a production interface.

​
Embed Core and Baseline

Add both packages when an application needs the public engine and its admitted offline resource set:
Register the baseline package explicitly with AddPidgeonBaselineData() in your dependency-injection composition. Core does not silently reach into the legacy data tree, and an absent required data package degrades with a typed result rather than pretending the resource exists.

​
Connect an MCP client

Use the governed adapter package:
Do not substitute @pidgeonhealth/mcp. That name is a contained legacy package and is not part of the governed public beta release train.
The adapter shells out to the local CLI in CLI mode. It does not turn the MCP client into an autonomous healthcare operator: the human remains responsible for scope, review, and use of the resulting evidence.

​
Verify before expanding

  1. Pin the package versions in source control.
  2. Run a synthetic example without patient data.
  3. Review the generated output and validation result.
  4. Record the command, version, seed, and result when repeatability matters.
  5. Report a reproducible issue through the relevant Pidgeon Health repository.
Next, read the test-data field guide or the human-and-agent workflow map.