TechnicalResearch

How I Upgrade My Hermes Agent Team 🧠 Bot Mode, Skills & MCP Tips

Wanderloots21m 40s

Callum demonstrates how to upgrade a specialist researcher agent in Hermes by equipping it with custom skills, access to a Zotero reference library, and integration with Consensus for academic paper discovery. The upgraded researcher then collaborates with an orchestrator and librarian agent to produce high-quality research reports on agentic memory poisoning, showcasing how improving individual agents enhances team performance.

Summary

Callum introduces a systematic approach to upgrading specialist agents within the Hermes bot framework, focusing on the researcher agent as a foundational team member. He explains that upgrading individual agents improves overall team quality, and the process is applicable to any agent specialization or external systems.

The upgrade involves three main components. First, an LLM wiki research skill is installed that provides multiple research lenses: balanced research (mixing primary and secondary sources), academic research (prioritizing papers and official materials via Consensus), current reality (official documentation), and combined research (running multiple passes and mapping sources). This skill integrates with a built-in grounded citations capability for proper footnote tracking.

Second, Zotero integration is enabled through a read-only skill that allows the researcher to query an existing personal research library. The connection uses deterministic tooling to validate information from Zotero, save copies to a research workspace, and present usable evidence. Callum notes this could be expanded to write access in the future, and mentions the synergy between Zotero and Obsidian for tracking quotes and linking highlights.

Third, Consensus is integrated via Model Context Protocol (MCP) for bounded academic paper discovery. This provides access to quality academic sources, with metadata preservation and requirements to follow original links before making claims. The MCP connection allows direct searching without modifying the knowledge base.

The upgraded researcher is then tested within a group chat alongside an orchestrator (quality assurance manager) and librarian (knowledge management system). When tasked with a combined research pass on agentic memory poisoning, the researcher successfully queries both Zotero and Consensus, returns citations with grounded evidence, and presents findings to the orchestrator for review. The orchestrator requests minor revisions, which the user approves, then hands off to the librarian.

The librarian's review skill ingests the report into an LLM wiki, creating new concepts and connections to existing entries. This demonstrates the synergistic workflow where researcher quality improvements propagate through the team. Callum notes the current process involves some manual advancement between agents but indicates future work will use Hermes' kanban feature for more consistent, automated handoffs and the ability to run researchers on scheduled loops.

About this episode

What if updating ONE bot made your entire agent team better? 👀 In today's video, I'll walk through how to enhance your bots in the NEW Hermes Bot mode to boost the quality of the whole team ✨ Try Consensus MCP Free: https://get.consensus.app/wanderloots26 Hermes now has bot mode, which enables effortless agent-to-agent communication. Bot mode brings in Hermes profiles, an existing feature, but makes it easier to work with multiple bots simultaneously. You can even put them in group chats with each other to coordinate on specific tasks. But creating the team is just the beginning. It's important to customize & optimize each bot in the team so that it can contribute what it needs to for the enhanced benefit of the entire workflow & team. To test that out, in this video, I give the Researcher agent (bot) 3 new skills: 1) improved research methodologies 2) connection to existing research systems with Zotero 3) discovery of new quality research (academic & papers) with Consensus While I use the Researcher for today's example, this process works for any specialized bot & agent. My hope is that this video helps you not only understand how to use bots, but also the bigger picture of what this means for working with agents across any workflow. For more on Hermes: Hermes Agent, by Nous Research (https://hermes-agent.nousresearch.com/), is a self-evolving AI Agent harness. That means it has skills, tools, and memories that enable the agent to tap into past conversations, codify style & preferences, & build custom workflows that improve every time you use them. Hermes has persistent memory, which means it extracts key information and remembers it over time, building a profile of you & how you work. These memories are used to build skills that improve over time as you repeat your workflows. Over time, you'll need to ask Hermes to do less, while still achieving more (without repeating yourself). Hermes is free & open source, so you can be confident that it is an agent you can use for the long haul. You can easily connect to local models like Gemma4 or Qwen through Ollama, so you can run your agent privately & freely. The new desktop app makes it easier than ever to get started. Whether you're new to agentic AI or are confident leveraging agent harnesses, this tutorial walks through everything you need to know to get Hermes working for you. For more on connecting to Ollama: https://www.youtube.com/watch?v=4KXLW9Y1r4c I also touch on why the skills & architecture of Hermes make Obsidian & the LLM Wiki a natural companion of this incredible agentic AI tool. More on LLM Wiki here: https://youtu.be/n4EVksU_EOs I hope you enjoy! ✨ P.S. I greatly appreciate any feedback, please let me know what you think 😊 Join My Membership: YouTube https://www.youtube.com/channel/UCFiU1vIpPD3lQltke_18m3A/join Patreon: https://www.patreon.com/c/wanderloots 💌 Sign up for my [free newsletter: Recalibrating](https://paragraph.xyz/@wanderloots.eth?referrer=wanderloots.eth) 🏡 Wander my Digital Garden https://wanderloots.xyz Timestamps: 00:00 Are More Bots Better? 00:47 Today's Goals & Outline 01:01 How Skills Can Level Up Agents 02:21 My Agent Team: Structure & Loop 04:27 Installing 3 New Skills 04:50 Enhanced Research Skill 06:58 Why Zotero? 07:35 The Zotero Research Skill 08:39 Connecting Zotero To Your Agent 09:26 Testing Zotero Skill & Connection 10:38 Academic Source Discovery Skill 11:20 Why Connect To Consensus MCP? 12:24 Adding MCP Tools To Hermes 14:21 Testing All 3 Skills Together 15:51 Reviewing The Research Report 17:32 Orchestrator's Quality Check 17:56 Adding To LLM Wiki (Knowledge Base) 20:22 Next Steps: Structured Workflows & Optimizations LINKS (MY WORLD) 🧭 [Recalibrating Newsletter Home](https://paragraph.xyz/@wanderloots.eth?referrer=wanderloots.eth) 🏔️To start reading from the beginning: [Recalibrating Newsletter Entry 1: What Recalibrating Means To Me](https://wanderloots.substack.com/p/1-what-recalibrating-means-to-me) 🌍 My [Website](https://wanderloots.com/) 📸 My [Print Shop](https://wanderloots.darkroom.com/) ✨ SOCIALS 🟣 [Farcaster](https://warpcast.com/wanderloots.eth) 📸 [Instagram](https://www.instagram.com/_wanderloots/) 📰 [Flipboard](https://flipboard.com/@_wanderloots) 📍 [Pinterest](https://www.pinterest.ca/wanderloots/) 🐦 [X (Twitter)](https://twitter.com/_wanderloots) 🤖 [Reddit](https://www.reddit.com/user/_wanderloots) MY FAVOURITE TOOLS 😴 🤯 The [Waking Up App](https://dynamic.wakingup.com/guestpass/SC4914439) (use this link for a 30 day free trial) 📝 [Obsidian](https://obsidian.md/) (decentralized note-taking) 📹 [Adobe Suite](https://prf.hn/l/lQ9DwpA) (general creativity) ❤️ [Welltory](https://app.welltory.com/payments/plans/main/?coupon=wanderloots) (Health Tracker) EQUIPMENT USED 6. Camera [Sony A7iii](https://amzn.to/3seSHv6) 7. Lens [Sony F2 28 mm](https://amzn.to/3TiWCT2) 8. Tripod [K&F Concept](https://amzn.to/3soCKCP)

Key Insights

  • Callum argues that upgrading foundational team members like the researcher first creates quality improvements that propagate to the entire agent team, rather than simply adding more bots
  • The LLM wiki research skill provides four distinct research lenses that can be deployed individually or combined—academic passes map to current reality to bridge peer-reviewed sources with state-of-the-art practice
  • Deterministic tooling in the Zotero skill validates information from the reference manager to ensure alignment before presenting evidence, treating it as a safety mechanism against library corruption
  • The combined researcher report found more relevant papers by searching for broader concepts like 'memory' and filtering by harm, rather than direct keyword matches like 'poison'—demonstrating semantic search effectiveness
  • The librarian's automatic connection of research findings to existing wiki concepts (agentic memory poisoning linking to multi-agent workflows and agentic memory) shows how structured knowledge bases enable emergent concept relationships

Topics

Hermes bot framework and specialist agentsResearcher agent skill customization and capabilitiesZotero reference manager integrationConsensus academic discovery and MCP integrationMulti-agent team workflows with orchestrator and librarianGrounded citations and knowledge managementAgentic memory poisoning research case studyQuality assurance and review processes between agentsLLM wiki knowledge base systemFuture automation with kanban workflows

Transcript

[0:00] And that's it. These skills are now installed and available to this researcher agent. So, this did a better job than just asking it to look for poison because I asked it to search for memory and then limit the results to paper that harm memory. Hermes bot mode lets you create specialized agents that improve your tasks both solo and as part of a team of agents. But more bots doesn't necessarily mean better. Today, I want to level up one specialist agent to see what it can do and how it can boost the rest of the team both now and in the future. Hi, my name is Callum, also known as Waterloots, and welcome to [0:31]…

Full transcript available for MurmurCast members

Sign Up to Access

More from Wanderloots

Get AI summaries like this delivered to your inbox daily

Get AI summaries delivered to your inbox

MurmurCast summarizes your YouTube channels, podcasts, and newsletters into one daily email digest.