Async Hermes Agent Documentation

Async Hermes Agent is an async-only Python library derived from the upstream NousResearch/Hermes Agent, based on release v2026.8.18. It converts the Hermes Agent harness into coroutine entry points, retaining the upstream agent loop, tools, providers, MCP, skills, memory, sessions, trajectories, runner, and batch runner. The library supports Python 3.11 through 3.13 and is installable from PyPI. It maintains familiar module paths and public names, with I/O-bearing calls awaited, allowing the host application to own its HTTP service or UI while the agent remains a reusable harness.

The library provides an async library surface where model, tool, MCP, session, trajectory, runner, and batch-runner entry points are awaited under their upstream names. It includes extensible tool use with retained terminal, file, web, browser, vision, skills, memory, session-search, clarification, delegation, and plugin tools. MCP support enables dynamic stdio, Streamable HTTP, and SSE server discovery with ordered tool observations and explicit async lifecycle cleanup. Persistent context is managed through SQLite sessions, memory, user-profile context, FTS history search, checkpoints, and resume.

For training data, the library preserves reasoning, tool calls, observations, and final answers, and can generate JSONL trajectories with bounded batch concurrency. It is service-ready, allowing embedding in FastAPI or another async host without bundling a second service framework or HTTP contract. The implementation uses coroutine transports for provider, network, MCP, and subprocess paths, while regular-file operations use aiofiles and SQLite uses aiosqlite, promising directly awaitable, event-loop-nonblocking entry points.