headroomlabs-ai/headroom
Headroom is a context compression layer for AI agents, reducing token usage by 60-95% across tool outputs, logs, RAG chunks, files, and conversation history, while maintaining accuracy.
Awesome AI Agents › End-to-End Optimizers
End-to-End Optimizers
topoteretes/cognee
Cognee is an open-source project that provides scalable and modular memory solutions for AI agents using ECL pipelines, enabling efficient data ingestion, knowledge graph generation, and query capabilities to enhance AI performance and reduce hallucinations.
letta-ai/letta
Letta is an open-source framework for building stateful AI agents with advanced reasoning, transparent long-term memory, and context management, supporting multiple LLM backends and providing a graphical development environment.
langchain-ai/opengpts
OpenGPTs is an open-source platform that replicates and extends OpenAI's GPTs and Assistants API, offering customizable cognitive architectures, extensive tooling, and support for multiple language models and vector databases.
aiwaves-cn/agents
Agents 2.0 is an open-source framework that enables the training and self-evolution of autonomous language agents through symbolic learning inspired by neural network methodologies.
getzep/zep
Zep is an AI memory foundation that uses a temporal Knowledge Graph to enable continuous learning and personalized experiences for AI agents by efficiently managing and retrieving contextual chat and data artifacts.
algorithmicsuperintelligence/optillm
OptiLLM is an optimizing inference proxy that significantly improves the accuracy and performance of large language models on reasoning tasks without requiring any model training, by applying over 20 advanced optimization techniques during inference.
mll-lab-nu/RAGEN
A reinforcement learning framework for training multi-turn reasoning LLM agents on the StarPO algorithm, shipping ten Gym-compatible environments plus diagnostics for echo-trap and reasoning-collapse failure modes.
langfengQ/verl-agent
verl-agent is an extension of veRL designed for scalable training of large language and vision-language model agents via reinforcement learning, featuring customizable input structures, diverse RL algorithms including GiGPO, and support for multi-modal environments.
lupantech/AgentFlow
Trainable tool-integrated agentic framework that splits reasoning into Planner, Executor, Verifier and Generator modules and optimises the planner online inside the running system with the Flow-GRPO algorithm.
stanford-iris-lab/meta-harness
Reference framework from the Stanford IRIS Lab for automated search over task-specific model harnesses, the code around a fixed base model that decides what to store, retrieve and show while it works.
evo-hq/evo
Autoresearch orchestrator that discovers what to measure in a codebase, instruments the benchmark, then improves the code by tree search over experiments run by parallel subagents in isolated git worktrees.
ysymyth/awesome-language-agents
A curated list and comprehensive resource of language agents based on the Cognitive Architectures for Language Agents (CoALA) framework, including extensive research papers and references.
KhoomeiK/LlamaGym
LlamaGym is a framework that simplifies fine-tuning large language model agents using online reinforcement learning in Gym-style environments.
agiresearch/A-mem
A-MEM is an advanced agentic memory system for LLM agents that dynamically organizes, links, and evolves memories using intelligent indexing and semantic search to enhance agent performance on complex tasks.
RUC-NLPIR/Arbor
Autonomous research agent that grows a hypothesis tree, edits code, runs real experiments in isolated git worktrees and keeps only the gains that survive held-out validation.
TIGER-AI-Lab/verl-tool
VerlTool extends the verl reinforcement learning framework with a unified tool-calling API, letting tool-using LLM agents be trained end to end with asynchronous multi-turn rollouts.
seanpixel/Teenage-AGI
Teenage-AGI is a Python project that creates an AI agent with persistent memory and internal thought processes using OpenAI and Pinecone, inspired by generative agents research.
DataBassGit/AgentForge
AgentForge is a low-code, extensible framework for rapidly developing, testing, and iterating AI-powered autonomous agents and cognitive architectures compatible with multiple large language models and local AI models.
NVlabs/ToolOrchestra
NVIDIA research release of ToolOrchestra, an end-to-end reinforcement learning method that trains small orchestrator models to coordinate tools and specialist LLMs across multi-turn agentic tasks.
microsoft/Trace
Trace is a PyTorch-like Python library that generalizes back-propagation to optimize AI agents end-to-end using general feedback such as rewards, losses, and natural language text.
agentscope-ai/Trinity-RFT
Trinity-RFT is a framework for reinforcement fine-tuning of LLMs and LLM agents, splitting the process into an Explorer that collects agent-environment experience, a Trainer, and a Buffer data pipeline.
FoundationAgents/AFlow
AFlow automatically generates and optimises agentic workflows by running a Monte Carlo tree search over a code-represented workflow space, using an LLM optimiser and a benchmark evaluator to replace hand-designed agent pipelines.
Gen-Verse/Open-AgentRL
Open-source reinforcement learning stack for training LLM agents, bundling the RLAnything, AutoTool and DemyAgent research projects with GRPO, PPO and RLHF training, multi-turn tool use and distributed runs.
Optima-CityU/LLM4AD_Next
LLM4AD_Next turns a natural-language problem description into a runnable evolutionary search project, using LLM agents to generate and iteratively evolve candidate algorithms through methods such as EoH, ReEvo and MCTS-AHD.
statelyai/agent
Stately Expert is a framework for building intelligent AI agents powered by state machines and enhanced with observations, feedback, and flexible decision-making policies, integrating multiple AI model providers via the Vercel AI SDK.
turing-machines/mentals-ai
Mentals AI is a tool for creating and operating AI agents using markdown files, featuring recursive loops, memory, and tool integration without traditional programming.
opennars/opennars
OpenNARS is an open-source general-purpose AI reasoning system based on Non-Axiomatic Reasoning System (NARS) designed to process tasks and learn adaptively under conditions of insufficient knowledge and resources.
jlin816/dynalang
Dynalang is an advanced agent that leverages diverse language types to predict the future and solve tasks using a multimodal world model, demonstrated across multiple embodied AI environments.
SwiftSage/SwiftSage
SwiftSage V2 is a generative agent system that combines fast and slow thinking processes using large language models and in-context reinforcement learning to solve complex reasoning tasks through iterative feedback and code execution.
zjunlp/KnowAgent
KnowAgent is a framework that enhances large language model-based agents' planning abilities by augmenting them with an extensive action knowledge base and iterative self-learning from generated action trajectories.
James4Ever0/agi_computer_control
Cybergod is an autonomous AGI program designed to perform any task to earn money independently, featuring advanced terminal interaction, reinforcement learning, and an integrated digital economy system.
xufangzhi/ENVISIONS
ENVISIONS is a neural-symbolic self-training framework that enhances large language models through environment-guided interactive evolution to improve reasoning and problem-solving abilities.