Skip to content
labs.Independent tools. Open source.
← All tools
AI & insights

Elrond

Compare answers from AI models. Run a debate and read the combined result.

Get started ↓View source code
Platform
macOS
Licence
MIT
Documentation
Read the guide
Elrond comparing two local model answers, a moderator verdict and a final synthesis
An example deliberation with two Ollama models, showing sample data in the actual app interface.01 / 02
A closer look

In the app

2 images
Elrond Agents dialog with provider assignments, two enabled agents and a synthesizer selector
02 / 02

Choose your agents

The actual app's agent configuration dialog, from the project documentation.

Model deliberation
Elrond comparing two local model answers, a moderator verdict and a final synthesis

An example deliberation with two Ollama models, showing sample data in the actual app interface.

1 / 2

Put a few minds on it

Elrond is a desktop workspace for asking the same question of multiple AI models. Compare their initial answers, run a moderated debate when useful, and read a final synthesis that brings the discussion together.

Supported options include OpenAI, Anthropic, Google and local models served through Ollama. The project also documents supported CLI-based provider connections.

What you can do

  • Choose models for individual agents and a synthesizer.
  • Compare responses side by side in one session.
  • Follow debate rounds and a moderator’s verdict.
  • Inspect per-turn statistics alongside the discussion.
  • Organize and revisit sessions in the desktop interface.

Getting started

git clone https://github.com/Captain-Sangam/Elrond.git elrond
cd elrond
make install
make export

Launch Elrond from Spotlight or Applications. Follow the setup wizard to configure providers and choose models. For debate, enable at least two agents and select a synthesizer.

You can start with a running Ollama server and models already downloaded. See the documentation for provider setup and development instructions.

Things to know

Elrond’s source code is free under the MIT licence. Cloud model usage may require paid API access or a supported subscription. Local Ollama models require suitable hardware and a running server.

Prompts sent to cloud providers leave your machine and are handled under those providers’ policies. Model agreement is useful context, but does not establish that an answer is correct.

Explore the source

Read the source code, review the installation guide, or contribute a useful improvement.