Agent Memory System is an open-source agent orchestration harness developed by RavByte-AI. It provides structured memory retention and retrieval mechanisms designed to help autonomous agents store, organize, and query state across long-running tasks and multi-turn workflows. In day-to-day use, the system acts as a backend memory infrastructure for agent pipelines. It implements structured storage mechanisms—including graph-oriented relational nodes and indexed recall—allowing agents to maintain execution context without overflowing LLM context windows. Developers interact with the system via code modules, integration APIs, and provided repository examples. Unlike traditional stateless coding harnesses that lose operational context between prompts or process terminations, Agent Memory System isolates memory management into a standalone subsystem. It focuses on persistent entity tracking, graph-based relationship mapping, and verifiable retrieval accuracy supported by built-in benchmark suites. The project is available as an open-source repository hosted on GitHub, catering to developers building and testing custom agent architectures.
Graph-based memory management — Organizes entities, execution steps, and interaction histories into structured relational graphs. - Built-in benchmarking suite — Includes testing tools to measure memory retrieval accuracy and context retention across runs. - Decoupled harness design — Separates the core memory store from specific execution loops for modular agent integration. - Integration examples — Provides scaffolded implementations showing how to connect memory stores to agent prompts. - Continuous integration tests — Features automated test suites to validate graph storage stability and retrieval integrity.
Use cases
Use Case: Long-Horizon Agent Tasks Developers can integrate the memory harness into autonomous agent pipelines to retain tool outputs, architectural plans, and file modifications across hundreds of execution steps.
Use Case: Cross-Session State Persistence Teams building CLI and automation agents can persist codebase knowledge and past developer feedback across independent execution sessions without reloading full contextual files.
Use Case: Agent Memory Evaluation AI engineers can run the repository's benchmark scripts to test and compare retrieval precision, graph stability, and token efficiency for various memory configurations.
Who it is for
This harness is suited for AI researchers, framework developers, and software engineers designing persistent autonomous agents that need stateful memory management. It is not intended for non-technical users or as a standalone out-of-the-box terminal coding assistant.
Install guide
Installation Clone the repository and install dependencies locally (standard Python setup): bash git clone https://github.com/RavByte-AI/agent-memory-system.git cd agent-memory-system pip install -e .
First Run Set your model provider API key if using integrated LLM retrieval modules: bash export OPENAI_API_KEY="your-api-key-here"
Example Session Run a sample harness script or execute the benchmark suite to verify the memory system: bash python examples/basic_usage.py python -m benchmarks.run_all
Data Olympus — Data Olympus is an open-source knowledge base and Model Context Protocol (MCP) server harness created by knaisoma. Built in Python…
m1nd — m1nd is an open-source, Rust-based code intelligence harness and Model Context Protocol (MCP) server created by maxkle1nz. Designed to provide…
GoodMemory — GoodMemory is an open-source agent memory orchestration harness and context management system developed by hjqcan. Designed to give AI agents…
Grite — Grite is an open-source terminal coding agent harness developed by Neul Labs. Implemented in Rust, it provides a lightweight command-line…
Mneme — Mneme is an open-source orchestration harness and memory governance framework developed by MnemeHQ designed to prevent architectural drift across the…
Flameox — Flameox is an open-source coding agent harness and orchestration framework maintained by morluto. Built to support hybrid development workflows across…