OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Multi-Agent Systems

Kadath

A Python-based evolutionary multi-agent runtime that breeds, evaluates and improves autonomous agents across reproducible epochs.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Kadath — OpenViST project profile in Multi-Agent Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when an agent task has a defined goal and a measurable way to evaluate outcomes across generations.

Who it is for

  • Researchers exploring evolutionary methods for LLM-based agents.
  • Developers optimizing prompts or complete agent frameworks against measurable goals.
  • Teams comparing populations of autonomous agents over repeated generations.

Documented uses

  • Evolve agent prompts and strategies against a measurable benchmark.
  • Compare autonomous agents through repeated population-based competition.
  • Test mutation, selection and reproduction strategies for multi-agent systems.
  • Maintain run evidence, benchmark locks and Git-based genome lineage.
Before you choose

Requirements & limits

Kadath is a public, open-source Apache-2.0 Python project intended for self-hosted use through a CLI on Linux, macOS or Windows. The documented first-launch setup requests an OpenAI API key and model ID.

  • A measurable benchmark is central to the workflow; the Architect must define supported measurements, evidence sources and scoring rules.
  • The system evolves complete agent frameworks, including prompts, implementations, tools, dependencies and supporting files.
  • The kernel keeps experimental control separate from the evolvable organisms and locks key run inputs before execution.
  • Runs can require substantial model-token use because the project evaluates populations over multiple generations.
  • The documented setup prepares PostgreSQL, MinIO, LiteLLM and SearXNG services locally.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Kadath for a first evolutionary run

Prepare the local runtime and define the inputs required to begin a Kadath run.

Before you start

  • A Linux, macOS or Windows environment with the Kadath project available.
  • An OpenAI API key.
  • An OpenAI model ID.
  • A goal, epoch duration, population size and epoch count.

Steps

  1. Launch ./kadath.sh to open the interactive terminal interface.

  2. On first launch, enter the OpenAI API key and model ID when prompted.

  3. Let Kadath prepare its PostgreSQL, MinIO, LiteLLM and SearXNG configuration and supporting services.

  4. Create a new run by providing the goal, epoch duration, population size and number of epochs.

  5. Review the Architect's benchmark approval screen before approving the run.

Expected outcome

The runtime is prepared and a proposed benchmarked run is ready for approval.

Guide limitations

  • The evidence documents setup and approval flow but does not provide a complete execution command sequence or model-provider compatibility beyond the OpenAI setup prompts.
  • The Architect must use measurable benchmarks and available runtime tools or external connectors; unverifiable outcomes require an honest measurable proxy.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
30 September 2026
Original website title
i3T4AN/KADATH · Official homepage title · retained in its source language
How we review sourcesSuggest a correction