MCP-Zero

SWE-smith creates Docker execution environments from GitHub repositories, synthesizes tasks such as file localization and program repair, filters instances that break one or more tests, and generates issue text. The resulting tasks and trajectories can support SWE-agent training or evaluation. It requires Docker and was developed for Ubuntu 22.04; Windows and macOS are not planned targets.

langgraph-bigtool stores tool names, descriptions, namespaces, and other metadata in a LangGraph Store so an agent can search for relevant tools before calling them. It is designed for registries containing hundreds or thousands of tools, supports in-memory and PostgreSQL backends, and lets developers replace the default retrieval logic with a custom function.

MagicSkills keeps reusable `SKILL.md` directories in one local pool and lets users compose named subsets for different agents or workflows. Runtimes that read `AGENTS.md` can receive a synchronized skill section, while agent frameworks can use the same collection through a command-line or Python tool interface. It is aimed at multi-agent projects that want to avoid copying and drifting skill folders, and requires Python 3.10–3.13 plus Git.

ToolRAG registers tools from MCP servers, embeds their descriptions, and uses semantic retrieval to select the functions relevant to a user query. The selected definitions can be passed in an OpenAI-compatible format and executed against the corresponding servers, while LibSQL stores the tool index and embeddings. It is aimed at assistants that need access to many APIs without sending every tool definition on every request.