[EPIC] Hermes Agent Capability Expansion — MCP, A2A, Local LLM, Memory, Computer Use, Voice #1120

New Issue

Timmy · 2026-04-07T21:17:05Z

Timmy commented

2026-04-07 21:17:05 +00:00

Epic: Hermes Agent Capability Expansion

Context:
Investigation of the AI-tools ecosystem revealed that the highest-value upgrades for our agent fleet are not more diffusion models, but agent infrastructure: interoperability protocols, local inference, persistent memory, desktop automation, and voice I/O.

Goal:
Equip Hermes agents with six foundational capabilities that move us from "chatbot with tools" to "sovereign, long-lived, multi-modal agent."

Scope:
This epic covers six workstreams. Each has its own issue with acceptance criteria, estimated effort, and assigned wizard (default: Bezalel for forge/infra items).

Workstreams:

#MCP — Native Model Context Protocol server/client integration
#A2A — Agent2Agent protocol for inter-wizard delegation
#LOCAL_LLM — llama.cpp backend for sovereign, offline inference
#MEMORY — MemPalace integration for cross-session agent memory
#COMPUTER_USE — Desktop/browser automation primitives (Claude Computer Use pattern)
#VOICE — TTS fallback via edge-tts for alerts and voice memos

Definition of Epic Done:

All 6 sub-issues either closed or explicitly descoped with reason
At least 3 capabilities are live in production on Beta/Alpha
Documentation updated in the-nexus/docs/hermes-capabilities.md
Demo recorded in Timmy Time showing cross-capability flow (e.g., agent remembers a preference → uses local LLM → speaks the result)

Owner: Bezalel
Target Date: 2026-05-31

# Epic: Hermes Agent Capability Expansion **Context:** Investigation of the AI-tools ecosystem revealed that the highest-value upgrades for our agent fleet are not more diffusion models, but **agent infrastructure**: interoperability protocols, local inference, persistent memory, desktop automation, and voice I/O. **Goal:** Equip Hermes agents with six foundational capabilities that move us from "chatbot with tools" to "sovereign, long-lived, multi-modal agent." **Scope:** This epic covers six workstreams. Each has its own issue with acceptance criteria, estimated effort, and assigned wizard (default: Bezalel for forge/infra items). **Workstreams:** 1. #MCP — Native Model Context Protocol server/client integration 2. #A2A — Agent2Agent protocol for inter-wizard delegation 3. #LOCAL_LLM — llama.cpp backend for sovereign, offline inference 4. #MEMORY — MemPalace integration for cross-session agent memory 5. #COMPUTER_USE — Desktop/browser automation primitives (Claude Computer Use pattern) 6. #VOICE — TTS fallback via edge-tts for alerts and voice memos **Definition of Epic Done:** - [ ] All 6 sub-issues either closed or explicitly descoped with reason - [ ] At least 3 capabilities are live in production on Beta/Alpha - [ ] Documentation updated in `the-nexus/docs/hermes-capabilities.md` - [ ] Demo recorded in Timmy Time showing cross-capability flow (e.g., agent remembers a preference → uses local LLM → speaks the result) **Owner:** Bezalel **Target Date:** 2026-05-31

Timmy referenced this issue

2026-04-07 21:17:06 +00:00

[MCP] Integrate Model Context Protocol into Hermes — client + server #1121

Timmy referenced this issue

2026-04-07 21:17:07 +00:00

[A2A] Implement Agent2Agent Protocol for Fleet-Wizard Delegation #1122

Timmy referenced this issue

2026-04-07 21:17:08 +00:00

[LOCAL-LLM] Standardize llama.cpp Backend for Sovereign Inference #1123

Timmy referenced this issue

2026-04-07 21:17:09 +00:00

[MEMORY] Integrate MemPalace for Cross-Session Agent Memory #1124