OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Evaluation & Observability

Proteus

Proteus is a Python framework that connects agent harnesses and models to an iterative self-evolution loop, allowing the harness to be rewritten across context-fresh episodes and measuring how its structure and behaviour change under specified goals or without a goal.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Proteus — OpenViST project profile in Agent Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Proteus when the object of study is how an agent harness changes over time, rather than only whether an agent completes a task or achieves a benchmark score.

Who it is for

  • Researchers studying recursive self-improvement and agent-harness evolution.
  • Agent builders comparing harness designs across repeated episodes.
  • Teams creating benchmarks or regression tracking for evolving harness versions.

Documented uses

  • Run repeated evolution episodes for an agent harness.
  • Compare structural changes between harness states.
  • Measure behavioural changes and action preferences.
  • Test goal-based, multi-goal and no-goal evolution conditions.
  • Connect a custom harness through a HarnessAdapter.
Before you choose

Requirements & limits

The repository is public, MIT-licensed Python software. The documented requirement is Python 3.10 or newer. The minimal harness runs offline; other adapters may require an API key, Docker or a research checkout.

  • The minimal adapter supports an offline pipeline without a model API key or Docker.
  • A custom harness requires an adapter, although the README describes no registration step.
  • Model-backed and third-party harness adapters have different documented prerequisites.
  • Proteus measures harness structure and behaviour, including no-goal conditions, rather than reporting only task pass-rates.
  • The repository describes the software as a research preview.

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

Run the offline minimal-harness measurement demo

Run the documented offline pipeline and measure changes in the bundled minimal harness.

Before you start

  • Python 3.10 or newer.
  • The Proteus package installed from the documented package name.
  • No model API key or Docker; the minimal harness runs offline.

Steps

  1. Install the package with the documented pip command.

  2. Run the minimal harness with the documented neutral, notes-review and tools-review arms, four seeds and eight episodes, writing results to the documented output directory.

  3. Run the documented measure command against that output directory.

Expected outcome

A run directory is produced and the measurement command reports harness-growth and behavioural-distance results for the configured arms.

Guide limitations

  • This recipe covers the documented minimal offline demonstration only; it does not configure a live model or a custom harness.
  • The example output is an illustration of the documented demo rather than a guarantee for another environment.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
3 October 2026
Original website title
proteus-evolve/Proteus · Official homepage title · retained in its source language
How we review sourcesSuggest a correction