BDD Pilot

Run Reqnroll and SpecFlow scenarios from VS Code and Cursor — one click, the right dotnet test filter, readable results.

The official Reqnroll extension targets Visual Studio. On VS Code and Cursor you're left hand-writing dotnet test --filter strings and scrolling raw console output to find which scenario failed. BDD Pilot gives you a Gherkin-shaped test tree, runs exactly what you click, and maps every result back to its scenario.

Why Pilot

• Run by domain, feature, scenario, @tag, or a single Scenario Outline row. Pilot builds the filter for you.
• See results on the scenario, not the test method. TRX and Cucumber JSON outcomes land on each .feature row, with the error snippet on hover.
• Know why it failed. Missing SDK, private NuGet feed, broken Playwright drivers, and pending steps are diagnosed in plain words.
• Switch environments safely. Pick STAGE (dev / test / stg / prod) from the status bar; stg and prod ask for confirmation, and prod stays blocked until you opt in.
• Hand the failure to your AI assistant. Copy a sanitized failure summary for Copilot or Cursor, or let the agent read it through the built-in MCP tools.

Quick start

1. Install BDD Pilot:
   • VS Code: Extensions view → BDD Pilot, or ext install anghelll.bdd-pilot.
   • Cursor, VSCodium, and other Open VSX editors: search BDD Pilot in Extensions (Open VSX listing).
2. Open a .NET solution with .feature files (Reqnroll or SpecFlow, any test framework that runs through dotnet test).
3. Open the BDD Pilot view in the Activity Bar and press Run on any row.

No project at hand? Open samples/minimal-bdd/ from this repo.

Who it's for

Pilot works alongside Test Explorer, not instead of it: Test Explorer for unit tests, Pilot for scenarios.

Features

Run what you click

• Tree view grouped by domain (Features/ folders) or by @tag, with pass / fail / skip and duration on every row.
• CodeLens Run / Debug above each Feature, Scenario, and Scenario Outline example row.
• Native Test Explorer integration with Run and Debug profiles.
• Search the tree by name, tag, or path, then Run filtered.
• Re-run failed, saved execution profiles, and Cancel that stops the whole dotnet / testhost / browser process tree.
• Copy effective command gives you the exact dotnet test … Pilot ran, ready for CI or a terminal.

Results you can act on

• TRX and Cucumber JSON outcomes decorate the tree and Test Explorer.

• Dashboard with run history, flaky scenarios (failure rate, average duration, last error), and the top diagnosis of the last run.

  

• Diagnostics for missing SDKs from global.json, NuGet feed and auth errors, vulnerability-as-error builds, filter mismatches, Playwright drivers, and pending or ambiguous steps.

• Evidence links to screenshots, traces, and videos when your suite produces them.

• Unmapped scenarios after a scoped run (in scope but missing from the TRX) are listed so you can jump to the .feature line.

Environments

The status bar hub shows Pilot · STAGE · mode · project. Click it to change stage, parallelism mode (debug / parallel / ci), or test project.

Optional config/.env.<stage> files are merged into the test process in memory only. Load order: config/.env.<stage> → config/.env.<stage>.local → config/.env.local. Start from config/env.example.

AI-ready failures

• Copy Failure Context for AI puts a structured, sanitized markdown summary of the last failed run on the clipboard. There is no embedded LLM.

• Agent mode (MCP): on VS Code 1.101+ and Cursor, the extension registers a BDD Pilot MCP server automatically. Tools:

  • pilot_discover_bdd — map features and scenarios
  • pilot_build_filter — build a dotnet test filter for a tag, feature, or scenario
  • pilot_failure_context — read the last failure (Pilot writes TestResults/bdd-pilot-last-failure.json after a failed run)

  Tool paths are restricted to the workspace root. To run the MCP server from a clone of this repo instead (Claude Desktop, CI), use npm run pilot:mcp with config/mcp.json.example.

Security

• Pilot never reads or stores credentials; secrets keep coming from your project's own mechanism.
• Everything written to the terminal, the clipboard, or MCP output is sanitized: passwords, tokens, JWTs, connection strings, AWS keys, and PEM keys are redacted.
• stg and prod runs require confirmation showing the test count and filter. prod is blocked until bddPilot.security.allowProductionRuns is enabled.
• No remote telemetry. Review copied context before sharing it outside your team.

Works well with

BDD Pilot is one piece of the ForgeOne BDD family. Each extension does one job.

Get Guardian and Format in one install with the BDD ForgeOne pack. With Guardian installed, Pilot can check bindings before a run (bddPilot.preRun.bindingGate).

Configuration

Pilot auto-detects the test project. Settings you're most likely to touch:

Same run in CI

Use the same STAGE and filter you run from the tree. BDD Pilot: Copy Effective Dotnet Command gives you the exact command line.

env:
  STAGE: test
steps:
  - uses: actions/setup-dotnet@v4
    with:
      dotnet-version: "8.0.x"
  - run: dotnet test samples/minimal-bdd/MinimalBdd.csproj --filter "Category=smoke"

Requirements

• VS Code 1.90+ or Cursor
• .NET SDK, with any NuGet feeds your project needs already reachable and authenticated

Learn more

• What's new
• Extension API for other extensions
• Roadmap
• Contributing and local build

Found a bug? Open an issue. MIT License.