infiAgent

Why we included this project

infiAgent is built for jobs that run for days: a research task, a long pipeline, something you cannot babysit. Instead of writing orchestration code, you assemble agents by editing YAML config files, and the framework manages context so it does not grow without bound as a task drags on. Interrupted runs resume from a checkpoint instead of starting over, which matters when a crash happens on day three. It also follows the Agent Skills open standard, so you can drop skill folders into a library and let agents discover and load them on demand. The runtime installs from PyPI and bundles the execution loop, tool executor, multi-provider LLM failover, and file-based persistent memory, making it a practical starting point for research-style agents or multi-agent setups without extra infrastructure.