Open Source LLM Agents
LLM agents are software systems that use large language models to interpret requests, plan steps, and take actions through tools or other services. They can handle tasks that require more than a single response, such as gathering information, coordinating workflows, or adapting to changing results. Agent designs vary in how they manage context, memory, tool use, and human oversight, so reliability and control are important considerations.
Open source tools in this area include agent frameworks, workflow runtimes, memory components, evaluation environments, and debugging utilities. When choosing one, consider its maturity, license, maintenance activity, security model, resource requirements, and compatibility with your models and integrations. These tools are useful to developers and researchers building, testing, or operating language-model-based assistants and automated workflows.
4 repositories · updated October 2, 2026
OrcaReplay: Time Travel for AI Agents, Debugging and Evaluation
OrcaReplay introduces "time travel" capabilities for AI agents, allowing developers to record, replay, fork, and debug any agent run with any model. It addresses the challenges of AI agent debugging by providing byte-for-byte reproducibility, offline analysis, and the ability to compare different models from specific checkpoints. This tool, built by the OrcaRouter.ai team, enhances observability and control over complex agent behaviors.
tooltrim: Drastically Reduce LLM Agent Tool Output Tokens, Improve Accuracy
tooltrim provides drop-in compression for LLM agent tool outputs, drastically cutting tokens while often improving answer accuracy. This provider-agnostic solution offers content-aware compression, faithfulness benchmarks, and seamless integration with popular frameworks or as an OpenAI-compatible proxy.
Declarative Agents: Profile-Driven LLM Agent Runtime in Go
Declarative Agents by Nokia Bell Labs offers a profile-driven runtime and design patterns for building tool-augmented LLM agents. It allows defining agents, their tools, states, and transitions via YAML profiles, eliminating the need for code changes for workflow alterations. This Go-based framework promotes flexible and dependable agent development.
Awesome Dynamic Agent Skills: A Curated List for LLM Agent Skill Systems
Awesome Dynamic Agent Skills is a comprehensive curated reading list accompanying a TMLR 2026 survey on dynamic, self-evolving skill systems for LLM agents. It provides a unified taxonomy, an eight-stage lifecycle, and a ten-operator vocabulary for understanding how LLM agents acquire and manage skills. The repository audits 124 papers, offering valuable insights into this rapidly evolving field.