Sparkth

Sparkth is a free, open source, extensible, science-driven, AI-first learning platform. It is under active development by Edly.

With Sparkth, users can create create courses from the chat UI of AI providers such as Claude or ChatGPT. Integration with these generative AI providers is achieved by the Model Context Protocol (MCP) standard.

Features:

• Filter/Event API for extensive customisation of Sparkth.
• MCP endpoints that make it possible to use Sparkth as an external tool in Claude/ChatGPT/Gemini.
• Course generation prompt template that follows good instructional design principles.
• Synchronization of generated course content with 3rd-party learning management systems (LMS), such as Canvas.

Tool Development Guide:

The tool development guide can be found here.

Roadmap:

• Native Rust API for integration of Sparkth into other applications.
• Development of a new webassembly-based standard for the creation of next-gen learning experiences.

Installation

Pre-built binaries

We provides binaries for macOS (x86_64 & aarch64), Linux (x86_64) and Windows (x86_64). They are published automatically for every version tag (vX.Y.Z) on GitHub under Releases. Each archive contains:

• Platform-specific binary (sparkth or sparkth.exe)
• .zip and .tar.gz formats
• SHA-256 checksums

Download from the CLI:

export sparkth_version=$(curl -s https://api.github.com/repos/edly-io/sparkth/releases/latest | sed -n 's/.*"tag_name": "\([^"]*\)".*/\1/p')

# Linux
curl -LO -o sparkth.tar.gz https://github.com/edly-io/sparkth/releases/download/$sparkth_version/sparkth-$sparkth_version-x86_64-unknown-linux-gnu.tar.gz
# macOS arm64
curl -LO -o sparkth.tar.gz https://github.com/edly-io/sparkth/releases/download/$sparkth_version/sparkth-$sparkth_version-aarch64-apple-darwin.tar.gz
# macOS x86_64
curl -LO -o sparkth.tar.gz https://github.com/edly-io/sparkth/releases/download/$sparkth_version/sparkth-$sparkth_version-x86_64-apple-darwin.tar.gz
# Windows
curl -LO -o sparkth.zip https://github.com/edly-io/sparkth/releases/download/$sparkth_version/sparkth-$sparkth_version-x86_64-pc-windows-msvc.zip

# Decompress
# For Linux/macOS:
tar -xzf sparkth.tar.gz
# For Windows:
unzip sparkth.zip

# Set execution mode (Linux/macOS only)
chmod a+x sparkth

# Smoke test
./sparkth --help

Building

Make sure that you have the following requirements:

• Rust (documentation):

  curl --proto '=https' --tlsv1.2 https://sh.rustup.rs -sSf | sh
  

• uvx, for MCP integration (documentation):

  curl -LsSf https://astral.sh/uv/install.sh | sh
  

Checkout the code:

git clone git@github.com:edly-io/sparkth.git
cd sparkth

Build the application:

cargo build --release

The compiled executable will be in target/release. Run Sparkth with:

cargo run --release --

Any CLI option supported by Sparkth can be appended to this command.

Docker

A pre-built Docker image for Sparkth is published automatically to the GitHub Container Registry on each push to the main branch.

Pull the ghcr.io/edly-io/sparkth:latest image with:

docker pull ghcr.io/edly-io/sparkth:latest

Alternatively, build the image manually with:

docker build -t ghcr.io/edly-io/sparkth:latest .

Then, run the Sparkth MCP server with:

# Expose MCP server to http://localhost:7727
docker run \
  --env CANVAS_API_URL=https://canvas.instructure.com/api/v1 \
  --env CANVAS_API_TOKEN=<your canvas API key> \
  --publish 127.0.0.1:7727:7727 \
  ghcr.io/edly-io/sparkth:latest

Any CLI option supported by Sparkth can be appended to this command.

Usage

Local usage

Add the sparkth (or sparkth.exe) executable as an external tool to Claude Desktop by editing the Claude configuration file:

# macOS
~/Library/Application\ Support/Claude/claude_desktop_config.json
# Windows
%APPDATA%\Claude\claude_desktop_config.json
# Linux
~/.config/Claude/claude_desktop_config.json

Add the Sparkth server, such that your configuration looks like the following:

{
  "mcpServers": {
    "Sparkth stdio": {
      "command": "/<PATH TO SPARKTH REPOSITORY>/target/release/sparkth",
      "args": ["--mode=stdio"],
      "env": {
        "CANVAS_API_URL": "https://canvas.instructure.com/api/v1",
        "CANVAS_API_TOKEN": "<your canvas API key>"
      }
    }
  }
}

Restart Claude Desktop. Ensure that the "Sparkth stdio" tools appear in the "Search and tools" menu. Then start a new chat and generate a course:

Use Sparkth to generate a very short course (~1 hour) on the literary merits of Hamlet, by Shakespeare.

Sparkth will generate a prompt that will help Claude generate this course.

Production deployment

Deploying Sparkth in production involves the following steps:

1. Run Sparkth in SSE mode, either with Docker or a pre-built image.

2. Run an HTTPS proxy, for instance Caddy with self-generated SSL certificates.

3. Add the Sparkth MCP server in Claude: Settings 🠂 Connectors 🠂 Add custom connector:

   Name=Sparkth SSE Remote MCP server URL=https://<yourdomainname.com>/

Options

Transport mode: server-sent events (SSE) / stdio

Sparkth MCP server can run in two modes, selectable via the --mode flag:

Mode	Description
stdio	Communicates via standard input/output streams.
sse	Starts an HTTP server with Server-Sent Events.

The default is sse on host http://0.0.0.0:7727.

This starts an SSE server on the given host and port with:

• GET / for establishing the event stream.
• POST /message for sending messages.

The stdio transport mode is appropriate for running with Claude locally:

./sparkth --mode=stdio

Development

Build in development mode:

cargo build

Run tests:

cargo test

⚠️ Note that in stdio mode, you will need to restart Claude Desktop every time you recompile, otherwise changes will not be automatically picked up.

License

This project is licensed under the MIT License — see the LICENSE file for details.