Skip to main content

myPKA vs Hermes: an AI team, not just an AI agent

myPKA compared with Hermes, side by side. Hermes is one agent you raise over time; myPKA ships a pre-built, governed expert team on day one, with a real markdown knowledge base, a visual cockpit, and in v4.0 self-updating, model-agnostic, and…

myPKA vs Hermes: an AI team, not just an AI agent

Most tools hand you one agent and a long road to make it useful. myPKA ships a pre-built, governed core team on day one: Larry the orchestrator plus 5 named specialists, each with a real job contract and routing built in, and a growing library of membership Expansion Packs that add more specialists in one install. A real markdown knowledge base the whole team reads. A visual cockpit. Self-updating, model-agnostic, and self-improving.

The org you hire, not the agent you raise.

The team, day one: Larry (orchestrator), Nolan (hiring), Pax (research), Penn (journal and PKM), Mack (automation), Silas (database). Membership Expansion Packs add more: the App Developer Pack brings Felix (frontend), Vex (security), and Vera (QA); the Designer Pack brings Iris (design system), Charta (infographics), and Pixel (visuals). One download from the Expansion Packs page, one install sentence to Larry.

With Hermes you climb to the top. With myPKA you start there.

A single agent like Hermes earns its power one stage at a time: first memory, then skills, then integrations, then orchestration, and finally a dashboard to see what it built. It is a sound path, and it takes real time to climb. The difference is that myPKA ships you in at the top rung on day one.

The Hermes climb (you build up to it, level by level): L1 download and install. L2 memory (SOUL / MEMORY files). L3 skills plus per-task model. L4 integrations (connect your apps). L5 orchestration (an agent runs sub-agents). L6 scheduled and async work. L7 agentic OS plus dashboard.

The myPKA start line (all of it, on day one): L1 runs inside Claude Code. L2 full markdown PKM graph. L3 specialist contracts plus model tiers. L4 MCP integrations, your apps. L5 Larry orchestrates the team (default). L6 tasks plus deliverables pipeline. L7 cockpit plus Deliverables folder (today).

The orchestration stage (an agent that manages sub-agents) is exactly what Larry does by default the moment you open myPKA. The final stage, a dashboard to see everything your agents built, is our visual cockpit plus a browsable Deliverables folder, shipping today.

A dashboard that shows you everything your agents actually produced? With a single-agent setup that is often pitched as the final, hard-won unlock. In myPKA it is standard: the visual cockpit plus a browsable Deliverables folder, available today.

Built on the ICOR(R) methodology

myPKA is not a random agent setup. It is the ICOR(R) methodology, the framework from Paperless Movement(R), turned into a working AI team. An organization with operating principles, not a single bot.

ICOR operating principles (how the team works). A governed team, not a lone agent. Clear roles, clear handoffs, and a paper trail behind every move. Larry the orchestrator plus named specialists working by clear Team Communications, routing, and handoffs. A Single Source of Truth rule: every fact lives in exactly one file, everything else links to it, never duplicated. Each team member keeps its own journal of durable insights, carried across sessions. Session logging so every working session is recorded. Good Documentation Practice (GDP): the quality discipline from regulated industries like pharma and medical devices, where the golden rule is simple, if it is not written down, it did not happen, so every record stays accurate, reliable, and reusable. Documented operating procedures: SOPs, Workstreams, and Guidelines.

The My Life concept (how knowledge is organized). Your knowledge base is structured the way you actually think, connected by links rather than buried in folders. My Life areas: Goals, Habits, Topics, Projects, and Key Elements. A CRM for the People and Organizations that matter to you. A Documents library for the records you keep. All connected by links, not folder depth, so nothing gets lost. The whole team reads and curates this one connected knowledge graph.

Task Management like a Pro (how the Planner works). The cockpit Planner follows the Task Management like a Pro approach, a Paperless Movement(R) method. Tasks are managed like a pro, not just listed on a flat to-do. A real method behind the Planner, the same one Paperless Movement teaches. Work moves through clear states, from open to in progress to done. The Planner lives inside the cockpit, alongside the team and the knowledge base.

ICOR(R) and Paperless Movement(R) are the proven framework underneath. myPKA is that framework, running.

Runs on your own Claude. Clean by design. Fully local.

myPKA works on the Claude plan you already have, stays inside Anthropic's authorized first-party app, and keeps every byte on your own machine.

No API key. No metered billing. Works on your Free, Pro, or Max Claude subscription with no Anthropic API key required, because it runs inside Claude Code, Anthropic's own authorized first-party app. You are not paying per token. You use the Claude plan you already have.

No impersonation. No account risk. Some competitors ride your Max login by impersonating the Claude Code client, which violates Anthropic's published terms and can get your account suspended without notice. myPKA never does that. It is content running inside the real, authorized client.

Nothing leaves your computer. Everything is markdown on your own machine. Nothing is uploaded, nothing is synced to us. Your data is yours.

ToS claim sourced to Anthropic's published usage terms. Stated factually, no straw man.

The full comparison

Hermes is excellent at what it does, and where it genuinely wins we say so. The difference is the shape of the bet: more reach versus a governed expert team with a real knowledge base.

Dimension

Hermes

myPKA (advantage)

Out-of-the-box team

One agent. The team is do-it-yourself: you clone "profiles" and wire them together over time.

(star) A governed core team of 6 named specialists, each with a real job contract and routing, working as a team from day one. Ready-made membership Expansion Packs add more specialists in one install.

See your outputs

Mostly lives in the chat thread. A dashboard to see everything is framed as the final "Level 7" unlock.

(star) A visual cockpit plus a Deliverables folder of browsable HTML, PDF, and docs. Standard, not a far-off level.

Updates

Git-pull the engine. Clean for them, because the product is code you re-pull.

(star) One-command self-update with a plain-English plan that never touches your data, plus a separate safe path for the cockpit interface.

Knowledge base

SOUL / MEMORY files plus a skills library.

(star) A full markdown PKM graph (journal, CRM, projects, documents) the whole team reads and curates.

Model choice

Model-agnostic. OpenRouter, a model picked per skill.

(star) No API key needed on your Claude plan is the baseline. Optionally, each specialist can run its own model: a portable model: tier field (reasoning / balanced / fast) plus OpenRouter with your own key.

Runs on your Claude plan

Rides your Max login by impersonating Claude Code. This violates Anthropic's published terms and can put your account at risk of suspension.

(star) Runs inside Claude Code, the authorized first-party app. Clean on your own Max plan, by design.

Privacy (fair tie, both sides)

Local, bring-your-own-key. Nothing leaves your machine.

(=) Local, bring-your-own-key. Nothing leaves your machine either.

Self-improvement

The agent auto-creates new skills for itself.

(star) A governed team-retro loop that proposes improvements for your approval. It never silently rewrites itself.

Hermes column stated fairly, no straw man. ToS claim sourced to Anthropic's published usage terms.

What is new in v4.0

Five pillars that close the gap on everything a single-agent tool can do, while keeping the team and the governance that make myPKA different.

  1. One-command self-updates. Pull the latest team and engine with a single command and a plain-English plan you approve first. It never touches your data.

  2. Per-specialist models, optional. The default is everything runs on your Claude plan, no API key. As an optional power feature, set each specialist's own model: a portable tier (reasoning / balanced / fast) plus OpenRouter on your own key.

  3. A self-improving team. A governed retro loop where the team proposes upgrades to itself. You stay in control. Nothing changes without your yes.

  4. Clean on Claude Max. Runs inside Claude Code, the first-party app, so it is ToS-safe by design on your own Max plan. No impersonation, no account risk.

  5. Portable across tools. The same team and knowledge base run across Claude Code, Codex, and Cursor. Your org is not locked to one host.

Same team. More power. Self-updating, model-agnostic, self-improving, ToS-safe, and portable. The governance and the expert team stay exactly where they were: yours.

Where Hermes shines

To be fair to Hermes: it has impressive reach into 20-plus chat apps and a large, active community skill ecosystem. If your priority is connecting one agent to many channels, that is real and worth respecting. myPKA's bet is different: a governed expert team and a real knowledge base, not more channels. Pick the shape that fits how you actually work.

The bottom line

Hermes is a brilliant single agent you have to raise. myPKA is the org you hire. A governed expert team, a real knowledge base, a visual cockpit, and now self-updating, model-agnostic, and self-improving. Day one.

Did this answer your question?