Published: 2026-06-03

Hermes Agent FAQ: Multitasking, Parallel Agents, Goals Mode, and Memory

Chapters / key moments (click to jump — plays here on the page)

Julian Goldie answers frequently asked questions about Hermes agent capabilities: how to handle multiple simultaneous projects with the Kanban board, whether Hermes can run parallel tasks across multiple profiles, how goals mode enables autonomous long-running work, and why Obsidian is the recommended memory layer rather than native agent memory for persistent context storage.

Source video

"Hermes Agent Is INSANE (FREE!)" by Julian Goldie SEO — Watch on YouTube →

Key Takeaways

The Kanban board lets Hermes break a complex task (e.g. "build a website") into subtasks (SEO, UI design, blog section) and assign them as separate agent workstreams running simultaneously.
Multiple Hermes profiles can run in parallel — create a separate profile per project and run them concurrently for truly simultaneous multi-project execution.
Goals mode / persistent goals lets Hermes work autonomously on long-running objectives — give it a schedule or goal statement and it continues working without manual triggering.
Native Hermes memory is not reliable for long-term context storage — use Obsidian as an external memory vault via MCP: "Connect to my local Obsidian vault and write/update conversations there."
Obsidian works as a shared memory layer across all your AI agents — Hermes, OpenClaw, and others can all read and write to the same Obsidian vault, keeping context in sync.

Multi-Task Setup: Kanban Board

To run multiple tasks simultaneously in Hermes, use the Kanban board feature rather than a single chat prompt. Give Hermes a high-level goal and the board breaks it into subtasks automatically:

Open the Kanban view in Hermes Desktop or via the web dashboard
State your goal: "Go and build a website"
Hermes decomposes it into cards: Structure SEO content → Design UI → Add blog section → etc.
Each card can be assigned to a separate agent profile running in parallel

External Memory with Obsidian

To set up persistent memory that survives session resets:

Install Obsidian (free, local Markdown note app)
Connect Hermes to Obsidian via MCP: tell Hermes "Connect to my local Obsidian vault at [path]"
Instruct Hermes to write conversations and updates to the vault automatically
Other agents (OpenClaw, etc.) can be connected to the same vault — shared context across all agents

Related Hermes Guides

Hermes Long-Running Tasks & Scheduling — goals mode, scheduling, and autonomous workflows
Hermes Memory Architecture — persistent memory setup and best practices
Hermes Web Dashboard — Kanban board and session management
Hermes MCP Tools — connecting Obsidian and other tools via MCP

What you can actually set up from this

Extracted from the video's own transcript — the specifics the original summary left out.

Commands & Code Shown

hermes profile create <name>

Purpose: Create a separate Hermes profile (own config, memory, skills) — also creates a <name> command alias.

When to use: Running several independent agents on one machine or VPS.

hermes model

Purpose: Pick provider and model — including OAuth logins that avoid API keys: MiniMax (OAuth) and xAI Grok OAuth (SuperGrok / Premium+).

When to use: Switching models, or using a subscription instead of pay-per-token API billing.

Reproducible steps

Parallel work: use the Kanban board
Drop several tasks on the Hermes Kanban board; it triages them into sub-tasks and you can watch each move across the columns instead of juggling terminals.
Separate projects in Telegram with topics
Use Telegram forum topics to keep projects apart in one chat.
Long-running projects: skills plus memory
Save a repeated workflow as a skill so you stop re-prompting it, and keep project context in memory (he uses an Obsidian vault) so every session starts informed.

Gotchas

Same VPS or a separate one for a second agent is a preference, not a requirement — profiles isolate config and memory on one machine.
Free models on Nous Portal are time-limited promotions (he used Step 3.7 Flash, "free for 30 days" in June 2026); check what is currently free in hermes model.
His rule for replacing SaaS tools with an agent: if the tool owns data you cannot get elsewhere (backlink indexes, traffic estimates), the agent cannot replace it; if it only does work, it can.