hermes-agent (uudruid74 fork)

What hermes-agent (uudruid74 fork) does

This fork of NousResearch/hermes-agent calls itself Live Agent. It began in July 2026 as cosmetic patches, such as colored response headers and agent icons in the TUI, and grew into a system where agents wake each other on events instead of the user polling for results: a kanban task completes, a cron job returns, or a worker gets stuck. The README describes five profiles, each with its own role, running behind one gateway and one bot token, with kanban-based routing instead of a group chat.

A plan_tool requires an ordered list of steps with a success criterion, created and approved before any state-changing work. Plans live on the board, so they survive context compression. A two-phase advance adds review: a worker calls advance with a completed-step summary and waits, then the dispatcher, or the human for self-made plans, calls review approve or deny, and a denial must carry a reason of up to 1024 characters. The README presents this as closing the self-certification loop.

Key features

• Agents woken by events such as kanban completion, cron results and stuck workers
• plan_tool with ordered steps and a success criterion
• Plans stored on the board so they survive context compression
• Two-phase advance with dispatcher review: approve or deny with a reason
• Five role-based profiles behind one gateway and one bot token
• Colored response headers and agent icons in the TUI

When to use it

• Running a small team of role-based Hermes profiles coordinated through kanban
• Requiring a reviewer to approve each plan step before the worker continues
• Resuming a plan from a DM, topic or fresh session after the context is compressed

Who it is for: Advanced Hermes Agent users who run several profiles and want enforced plans and review between them.

How it fits with Hermes Agent

It is a fork of Hermes Agent itself, extending the kanban and profile features with its own customizations.

Note: It is a modified fork maintained with its owner's own customizations, so its behavior can differ from upstream NousResearch/hermes-agent.

FAQ