A powerful experiment harness for systematic research. Define experiments as code, track runs with full reproducibility, and monitor execution in real-time.

Get Started Documentation 
defmodule MyExperiment do
  use Athanor.Experiment

  @impl true
  def run(ctx) do
    config = Runtime.config(ctx)

    for i <- 1..config["iterations"] do
      Runtime.progress(ctx, i, config["iterations"])
      result = run_trial(i, config)
      Runtime.result(ctx, "trial_#{i}", result)
    end

    Runtime.complete(ctx)
  end
end

Built for Rigorous Research

📐

Code-Defined Experiments

Define experiments as Elixir modules with typed configuration schemas. Version control your methodology alongside your code.

🔄

Reproducible Runs

Separate experiment instances from runs. Configuration snapshots ensure you can always reproduce past results.

📊

Real-Time Monitoring

Watch experiments execute with live progress, streaming logs, and results—all through a responsive web interface.

🔌

Programmatic Access

Full MCP server integration with 15 tools. Automate experiment workflows or integrate with your existing toolchain.

⚡

Supervised Execution

Each run executes in isolated processes with proper supervision. Failures are contained, cancellation is clean.

🗄️

Persistent Results

Structured results and logs persist to PostgreSQL. Query, analyze, and export your data however you need.

Quick Start

1

Clone and Setup

git clone https://github.com/BinaryMuse/athanor.git
cd athanor
mix setup
2

Start the Server

iex -S mix phx.server
3

Create Your First Experiment

Visit localhost:4000 and create an instance from one of the example experiments, or write your own.