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
stanford-iris-lab/meta-harness
Meta-Harness is the reference implementation released with the paper Meta-Harness: End-to-End Optimization of Model Harnesses. The project treats the harness, meaning the scaffolding code around a fixed base model that decides what information is stored, retrieved and shown to the model during a task, as the object being optimised instead of the model weights. The repository contains the reusable framework, an onboarding flow for applying the method to a new domain, and two reference experiments from the paper. The first, under reference_examples/text_classification, searches over memory systems for a text classification task. The second, under reference_examples/terminal_bench_2, evolves an agent scaffold against Terminal-Bench 2.0; the optimised harness produced in the paper is published separately in the meta-harness-tbench2-artifact repository. Both examples are Python projects driven by uv, with a short quick start that runs a single optimisation iteration or a smoke evaluation on one task. To apply the method to a new domain, the README directs the user to point a coding assistant at ONBOARDING.md and work through a conversation that produces a domain_spec.md file describing how to implement Meta-Harness for that domain. The shipped examples assume Claude Code as the proposer agent that generates candidate harness modifications, and switching to another proposer means adapting the claude_wrapper.py scripts in either example; the main requirement is a wrapper that logs proposer interactions cleanly. The authors describe the release as a cleaned-up version of the research code that has been verified to run but not tested further. The README also lists community reimplementations and applications, including a legal benchmark study, a Python library and CLI, a Claude Code skill port, and a sandbox-based variant. It is intended for researchers and practitioners working on agent scaffolding and harness optimisation.
https://github.com/stanford-iris-lab/meta-harness
Also in 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.