2025

Cluedo Mystery Calls

Description

A modern reinterpretation of Cluedo built during BU C&I Research Week. Players interrogate suspects via Bland.AI voice calls, review evidence, and solve a timed murder mystery — demonstrating rapid prototyping with AI code assistance.

Cluedo Mystery Calls – AI-Powered Murder Mystery

During BU C&I Research Week, students were developing projects using Bland.AI voice technology. Several had previously expressed interest in seeing how staff would approach a similar challenge. This became the motivation for building Cluedo Calls — a browser-based murder mystery where the suspects are not just cards or static characters, but voice-responsive AI personalities you can call on your phone.

The goal was to test how far a small prototype could go in a short space of time using GitHub Copilot as an accelerant for development. The result was a functioning playable experience: the user plays detective, phones Cluedo characters, hears their alibis, connects clues, and ultimately identifies the murderer.

The core question driving the work was simple.
If board game deduction became interactive and vocal, how much more immersive could it be?

Why Build This?

The project explored several research-aligned themes:

Reimagining traditional narrative formats

Board games rely on inference and imagination. By adding AI-powered voice interaction, the intention was to shift the experience from reading a card to interrogating a living suspect, introducing uncertainty, tone, hesitation, lies, and personality.

Testing rapid-prototyping with AI coding tools

With only a few evenings available, Copilot became a development partner — generating boilerplate, scaffolding components, and handling repetitive logic. What remained human-led were the design decisions, narrative dynamics, and interaction pattern.

Investigating how voice AI changes gameplay

Instead of players simply selecting a suspect, they hold a conversation. They must listen, evaluate tone, challenge contradictions. The social complexity of deception becomes part of the game system.

Features (Expanded)

  • AI phone calls to suspects
    The game connects to Bland.AI’s voice system so players receive live calls from characters. These conversations are unscripted within guidance prompts, meaning suspects may respond differently across playthroughs.

  • Distinct voices for classic Cluedo characters
    Each character personality was built with its own vocal profile and behaviour instructions. Miss Scarlet is evasive, Colonel Mustard is proud, Professor Plum over-explains. This builds replayability and storytelling richness.

  • Evidence-driven investigation UI
    Instead of guesswork, players review evidence logs, item discovery, movement timelines, and alibi statements. Calls generate notes automatically so players can track inconsistencies and motives.

  • Investigation log + accusation moment
    Once confident, the player submits an accusation to the Chief Inspector. If correct, the case closes. If wrong — well — justice may not be served.

  • Timed gameplay for tension
    A deadline hangs over the investigation. Each call, reconsideration, and note consumes time, maintaining urgency and forcing prioritisation.

  • Stylised noir interface
    The UI adopts grainy textures, dark palettes, and detective-office typography — mirroring noir cinema and creating atmosphere from first click to final reveal.

How It Was Developed

Architecture & Front-End

The project uses React with TypeScript, chosen for component-based development and type safety. React Router structures the application into:

  • Case file home
  • Suspect directory
  • Evidence log
  • Accusation & outcome

This separation allowed incremental development without rebuilding the interface.

Styling & Visual Identity

TailwindCSS was used to move quickly — utility classes enabled rapid iteration without full design systems. The noir aesthetic emerged through:

  • Low-contrast palettes
  • Film-grain textures
  • Bold serif headings with typewriter-style body text

Bland.AI Integration

Each suspect was defined through:

  • Role profile
  • Voice preference
  • Murder-night history
  • Prompts for behaviour under interrogation

Calls trigger via API, and responses are conversational rather than strictly scripted.

Development Process

Work happened over several evenings:

  1. Scaffold React + routing
  2. Integrate Bland.AI and test outbound calls
  3. Design the UI and evidence structures
  4. Add personalities & alibi logic
  5. Refine flow into a full session experience

Copilot handled repetitive tasks (state wiring, components, Tailwind classes) but the creative and investigative framing remained hand-authored.

Requirements

  • Node.js (v16+)
  • npm or yarn
  • Bland.AI account, API key, and call credits
  • Verified phone number for suspect call handling

This prototype demonstrates how traditional game structures evolve when AI voice interaction becomes the primary mechanic. It is both a research artefact and a playable experience — proof that with modern tooling, even a solo developer with a limited window can build something atmospheric, interactive, and conceptually ambitious.