NousResearch/hermes-agent

The agent that grows with you

About the project

Hermes Agent is an AI assistant that learns from its use by creating and improving skills, retaining knowledge, searching past conversations, and adapting to its users. It is for people who want to run an agent through a terminal or messaging platforms while connecting it to different AI models and scheduled tasks.

This repository also configures its own agents. See what hermes-agent tells them →

244kStars on the repository
200Mods indexed here, across every type
todayLast push, which is what freshness is scored on
MITLicence, which decides whether bodies are shown

llava

73

NousResearch/hermes-agent

Skill Claude CodeCodex needs its repo

Vision-language chat: VQA, captioning, image dialogue.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 16 tokens original MIT

modal

74

NousResearch/hermes-agent

Skill Claude CodeCodex

Serverless GPU cloud for ML jobs and model APIs.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 13 tokens original MIT

huggingface-hub

75

NousResearch/hermes-agent

Skill Claude CodeCodex

HuggingFace hf CLI: search/download/upload models, datasets.

not rated 244k +2.1k today C SkillSpector: warn 20 tokens original MIT

NousResearch/hermes-agent

Skill Claude CodeCodex needs its repo

SAM: zero-shot image segmentation via points, boxes, masks.

not rated 244k +2.1k today C SkillSpector: warn 18 tokens original MIT

nemo-curator

77

NousResearch/hermes-agent

Skill Claude CodeCodex

Curate LLM training data: dedupe, filter, PII redaction.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 22 tokens original MIT

obliteratus

78

NousResearch/hermes-agent

Skill Claude CodeCodex needs its repo

OBLITERATUS: abliterate LLM refusals (diff-in-means).

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 21 tokens original MIT

peft

79

NousResearch/hermes-agent

Skill Claude CodeCodex

Fine-tune large LLMs with LoRA on limited GPU memory.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 18 tokens original MIT

pinecone

80

NousResearch/hermes-agent

Skill Claude CodeCodex

Managed vector DB for production RAG and search.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 13 tokens original MIT

pytorch-fsdp

81

NousResearch/hermes-agent

Skill Claude CodeCodex

Fully sharded data-parallel training for large models.

not rated 244k +2.1k today A Snyk: pass 17 tokens original MIT

pytorch-lightning

82

NousResearch/hermes-agent

Skill Claude CodeCodex

Clean training loops with built-in distributed support.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 14 tokens original MIT

qdrant

83

NousResearch/hermes-agent

Skill Claude CodeCodex

Vector search engine for production RAG systems.

not rated 244k +2.1k today A Snyk: failSkillSpector: warn 13 tokens original MIT

dspy

84

NousResearch/hermes-agent

Skill Claude CodeCodex

DSPy: declarative LM programs, auto-optimize prompts, RAG.

not rated 244k +2.1k today A Snyk: failSkillSpector: warn 19 tokens original MIT

saelens

85

NousResearch/hermes-agent

Skill Claude CodeCodex

Train sparse autoencoders to interpret model features.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 14 tokens original MIT

simpo

86

NousResearch/hermes-agent

Skill Claude CodeCodex needs its repo

Reference-free preference alignment, simpler than DPO.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 13 tokens original MIT

slime

87

NousResearch/hermes-agent

Skill Claude CodeCodex

RL post-training for LLMs with Megatron and SGLang.

not rated 244k +2.1k today A Snyk: passSkillSpector: warn 18 tokens original MIT

stable-diffusion

88

NousResearch/hermes-agent

Skill Claude CodeCodex

Text-to-image generation, inpainting, and img2img.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 18 tokens original MIT

tensorrt-llm

89

NousResearch/hermes-agent

Skill Claude CodeCodex

High-throughput LLM inference on NVIDIA GPUs.

not rated 244k +2.1k today A Snyk: passSkillSpector: warn 18 tokens original MIT

torchtitan

90

NousResearch/hermes-agent

Skill Claude CodeCodex needs its repo

Pretrain LLMs at scale with PyTorch 4D parallelism.

not rated 244k +2.1k today A Snyk: passSkillSpector: warn 20 tokens original MIT

axolotl

91

NousResearch/hermes-agent

Skill Claude CodeCodex needs its repo

Axolotl: YAML LLM fine-tuning (LoRA, DPO, GRPO).

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 24 tokens original MIT

trl-fine-tuning

92

NousResearch/hermes-agent

Skill Claude CodeCodex

TRL: SFT, DPO, GRPO, RLOO reward modeling for LLM RLHF.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 28 tokens original MIT

unsloth

93

NousResearch/hermes-agent

Skill Claude CodeCodex

Unsloth: 2-5x faster LoRA/QLoRA fine-tuning, less VRAM.

not rated 244k +2.1k today A Snyk: passSkillSpector: pass 28 tokens original MIT

whisper

94

NousResearch/hermes-agent

Skill Claude CodeCodex

Transcribe and translate speech in 99 languages.

not rated 244k +2.1k today B Snyk: passSkillSpector: warn 13 tokens original MIT

mpp-agent

95

NousResearch/hermes-agent

Skill Claude CodeCodex

Pay HTTP 402 APIs via Machine Payments Protocol (MPP).

not rated 244k +2.1k today C SkillSpector: warn 16 tokens original MIT

stripe-link-cli

96

NousResearch/hermes-agent

Skill Claude CodeCodex

Agent payments via Stripe Link — cards, SPT, approvals.

not rated 244k +2.1k today A Snyk: failSkillSpector: warn 17 tokens original MIT

At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: