OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Frameworks

Agentdescent

AgentDescent is a Python framework for evolving agent skills, prompts and harnesses through parallel asynchronous proposals and diff-based aggregation. Workers propose edits to a shared artifact, and an aggregator resolves conflicts, combines compatible changes and commits accepted versions.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Agentdescent — OpenViST project profile in Agent Frameworks.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use AgentDescent when you have an agent artifact, task data, a way to run the agent and a scoring rule, and want to explore multiple candidate changes in an iterative optimization loop.

Who it is for

  • Developers iterating on prompts, skills or harness configurations with task-based evaluation.
  • Researchers comparing candidate agent variants across repeated runs.

Documented uses

  • Generate candidate changes to an agent artifact in parallel.
  • Score agent behavior on tasks and use the results to guide further artifact changes.
  • Merge compatible edits and resolve conflicts in a shared, version-controlled library.
  • Compare evolved variants using held-out evaluation in a configured workflow.
Before you choose

Requirements & limits

The project is documented as open source under the MIT license. The library supports Python 3.9 or later.

  • The user supplies the dataset, execution function, scoring method and model for a provider-backed evolution run.
  • The artifact's key structure affects whether concurrent edits can be combined or conflict.
  • The offline demo runs without a model API key; provider-backed examples configure an OpenAI-compatible endpoint and read credentials from environment variables.
  • The aggregation step performs conflict resolution and acceptance rather than averaging edits.

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 evolution demo

Install AgentDescent and start its offline demo loop without a model API key.

Before you start

  • Python 3.9 or later.
  • A source checkout of the AgentDescent repository, or the published package version described in its installation instructions.

Steps

  1. Install the published package with the command below, once that version is published; for source development, clone the repository and run the command below.

    Documented command
    pip install "agentdescent>=0.5.1"
    Documented command
    pip install -e .

  2. Run agentdescent demo.

Expected outcome

The documented demo runs an offline evolution loop and prints a held-out reward and the learned artifact.

Guide limitations

  • The README describes the package command as applying once version 0.5.1 is published.
  • The demo does not require a model API key; research comparison examples are outside the wheel and have separate setup instructions.
Documentation 1 ↗
Evidence and transparency

Sources & review

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