AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompare
Home/
Prompt Engineering/
agentdescent
agentdescent logo

agentdescent

Active·★ 207·MIT·Updated 2026-10-22

AgentDescent is a parallel, asynchronous framework for evolving agent skills, prompts and harnesses, where N workers propose diffs to a shared artifact and a barrier-free aggregator merges them into a git-backed ledger. Merging uses conflict resolution, fusion and statistical acceptance rather than averaging, and the project publishes measured results and a paper.

agentdescent sits in the Prompt Engineering category, so it is usually better judged by workflow fit, deployment style, and team needs than by isolated features alone. Based on the available metadata, it leans most clearly toward N workers propose edits in parallel; one aggregator merges and commits them. A representative use case is Evolve an agent skill or prompt from a scored dataset. The listed license is MIT. GitHub stars are currently 205. Relevant tags include agent-framework, agents.

This write-up is based on the project's README and public repository data. We have not hand-tested the tool.

© 2026 AgentIndex.app|Built by a 10-year iOS Developer.
Best by use caseAboutContactAdvertisePrivacy PolicyTerms of UseQYSGitHubBuy me a coffee ☕

Browse by Category

MCP ServersCode AssistantWorkflow AutomationRAG / Knowledge BaseMulti-AgentBrowser AutomationLLM InfraDev ToolingObservability

Not affiliated with Anthropic, OpenAI or Microsoft.

Guides
#agent-framework#agents#ai-agents#asynchronous#evolutionary-algorithms#gradient-descent#llm#llm-agents
$ Install
$ pip install "agentdescent>=0.6.0"
↗ Visit site★ GitHub
01

Features

01N workers propose edits in parallel; one aggregator merges and commits them
02Five-stage merge pipeline with conflict resolution and statistical acceptance
03Git-backed ledger of evolved skills and prompts
04Single evolve() entry point with pluggable strategies
05Offline demo that runs without a model API key
02

Why choose it

+N workers propose edits in parallel; one aggregator merges and commits them
+Five-stage merge pipeline with conflict resolution and statistical acceptance
+Evolve an agent skill or prompt from a scored dataset
+Comes with a clearly listed MIT license.
03

Trade-offs

!Compatibility coverage appears narrow, so your target environment should be validated before adoption.
04

Compatibility

Python
Python 3.9+
Verified via docs
05

Quick start

1
$ pip install "agentdescent>=0.6.0"
06

Use cases

↳Evolve an agent skill or prompt from a scored dataset
↳Speed up self-improvement loops by merging parallel edits
↳Research parallel optimization of agent artifacts
07

Alternatives

⚡Top VS:
VSinitrunner★40→VSAntOmniEvo★1.1k→VSKiroCrew★4.5k→VS

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs04Compatibility05Quick start06Use cases
Observer
★1.7k
→
initrunner logo
initrunner★ 40
Define AI agent roles in YAML and run them anywhere: CLI, API server, or autonomous daemon
VSCompare→
AntOmniEvo logo
AntOmniEvo★ 1.1k
An auto-evolution framework that optimizes anything — your 7×24 team of algorithm engineers.
VSCompare→
KiroCrew logo
KiroCrew★ 4.5k
A persistent workspace for development work that self-improves and continues beyond one session.
VSCompare→
Observer logo
Observer★ 1.7k
Why observe computer if computer can observe for you
VSCompare→
Ornn logo
Ornn★ 20
Your all-in-one skills-as-a-service platform to manage your skills, auto-generate skills and use agent skills as simple as calling an API.
VSCompare→
agent-device logo
agent-device★ 5.1k
Mobile app automation and verification for AI coding agents. CLI, MCP server, and typed Node.js API for iOS, Android, HarmonyOS, TV, web, macOS, and Linux.
VSCompare→
agenticos logo
agenticos★ 114
Sovereign Agentic AI Layer: your company's AI agents on your servers, models and rules. Skills, context files, 5700+ MCP servers, budgets, approvals and an audit trail. Apache-2.0, built on Pydantic AI.
VSCompare→
ggui logo
ggui★ 43
The universal interface layer between AI agents and humans. Generate rich UIs on demand via MCP.
VSCompare→
See all alternatives →
07
Alternatives
Birfy /
agentdescent
↗

Gradient descent, but the parameters are agents — a parallel, asynchronous framework for self-evolving agents (skills, prompts, harnesses). Diffs are the gradients; the aggregator is the optimizer.

207MIT
Last commit1d ago
StatusActive
CategoryPrompt Engineering
Links
Documentation↗Discussion↗Issues↗Releases↗
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit