Hermes Agent v0.6.0 Ships Profiles, MCP Mode, Docker Container

What changed

Hermes Agent v0.6.0 (The Multi-Instance Release) landed March 30, 2026. Profiles let one Hermes install run many isolated instances with their own config, memory, sessions, skills, and gateway service (commands: hermes profile create, hermes -p [name], hermes profile export/import). A token-lock mechanism prevents two profiles from sharing a bot credential. MCP Server Mode (hermes mcp serve) exposes Hermes conversations and sessions to any MCP client over stdio or Streamable HTTP. An official Docker container supports CLI and gateway modes. An ordered fallback provider chain auto-fails over when the primary errors. Feishu/Lark and WeCom adapters land alongside Slack multi-workspace OAuth, Telegram webhook mode, and Exa as a search backend.

Why it matters

Profiles eliminate the parallel-install / version-drift problem that comes with running Hermes across multiple roles. MCP Server Mode flips Hermes from a consumer of MCP to also a producer, so its skill history becomes queryable context for the AI tools you already use day to day. Official Docker reduces VPS deploy friction from Python environment plus systemd plus dependency pinning down to a single docker run. Ordered fallback provider chains move resiliency from per-call retry logic into the framework itself.

What to watch for

Profiles migration is opt-in: existing single-install configs continue to work as a default profile, so upgrades are non-breaking. Validate the ordered fallback chain against an induced primary outage before relying on it in production. Confirm that any MCP clients (Claude Desktop, Cursor, VS Code, Windsurf) you intend to query Hermes from are on a version that supports the transports you expose.