Use-Case Shapes: Coding, Personal, and Platform Bot

+15 Mana ✨
Hermes Features Overview(DOCUMENTATION)

Introduction

Hermes Agent is general-purpose, but most teams adopt it for one of three jobs. Recognizing which shape your use case fits helps you make the right choices about provider, gateway, and skills early on, and saves you from configuring features you will not need.

Key Concepts

  • Coding assistant: Hermes lives in your terminal, drives file/terminal/web tools, and pairs with you on code. The TUI is your primary surface.
  • Personal assistant: Hermes is your always-on agent across messaging platforms, with a SOUL.md personality and a long memory.
  • Platform bot: Hermes wraps a specific workflow (support triage, content moderation, internal Q&A) and exposes it as an unattended bot on Slack, Discord, or a website.

Real World Context

A solo developer building a SaaS product probably wants the coding assistant shape: TUI sessions, file editing, terminal access, and a strong code-capable model. A founder who wants an AI that reads their email, drafts replies, schedules meetings, and chats over Telegram wants the personal assistant shape. A team that wants @bot what is the deploy status? to work in their Slack wants the platform bot shape. Same Hermes binary, three very different setups.

Deep Dive

The three shapes differ along three axes:

text
Shape              Surface         Memory weight   Provider profile
─────────────────  ─────────────   ─────────────   ─────────────────
Coding assistant   TUI             Project-level   Capability > cost
Personal asst.    Gateways        User-level      Cost-balanced
Platform bot       Gateways/API    Workflow-level  Reliability + speed

Coding assistant

Optimized for deep, interactive sessions. You typically pick a top-tier model because every minute of dev time saved is worth more than the marginal token cost. Memory matters less because the project's repository is your context; you reach for @file and @diff more than long-term recall.

Personal assistant

Optimized for ambient availability. You usually wire up one or two messaging gateways (Telegram, iMessage), enable voice mode, and lean on the long-term memory and SOUL.md personality. Cost matters more here because the agent runs continuously, so a balanced model and a credential pool are common.

Platform bot

Optimized for unattended reliability. The agent runs as a background service, often in Docker, and answers requests from many users. You configure tight tool allowlists, fallback chains, and rate limits, and frequently expose the agent through the OpenAI-compatible API server so other software can call it.

Common Pitfalls

  1. Trying to make one Hermes install fit all three shapes, It works, but you will fight conflicting configurations. Use profiles (covered in later courses) to keep coding, personal, and bot setups isolated.
  2. Picking the surface before the shape, "I want a Slack bot" is a means, not an end. Decide what job the bot does first; the gateway choice falls out of it.

Best Practices

  1. Name your shape early, Even a one-line note in your config ("This Hermes is my coding assistant") helps you say no to feature creep later.
  2. Match the model to the shape, Capability-heavy for coding, balanced for personal, reliability-tuned for bots.

Summary

  • Three common shapes: coding assistant, personal assistant, platform bot.
  • Each implies different surface, memory weight, and provider profile.
  • Naming the shape upfront keeps configuration coherent.
  • Profiles let one Hermes install host multiple shapes without conflict.

Code Examples

yaml
# A one-line note in your Hermes config that future-you will thank present-you for
shape: coding-assistant
notes: |
  This Hermes is my code-pairing agent.
  Capability matters more than per-token cost.
  No always-on gateways, TUI only.
✓ Completed