HERMES AGENT ECOSYSTEM HIT 188K STARS.
HERE ARE 5 REPOS BUILDING ON TOP OF IT
THAT YOU SHOULD KNOW ABOUT.
1. GBRAIN (22.8K stars)
by Garry Tan, CEO of Y Combinator.
persistent memory layer built specifically
for Hermes and OpenClaw deployments.
markdown-first. self-wiring knowledge graph.
zero LLM calls for entity extraction.
Tan runs it himself: 146K pages,
24K people, 66 autonomous cron jobs.
what it adds to Hermes:
→ unlimited long-term memory (no char cap)
→ typed entities (people, companies, relationships)
→ hybrid search (pgvector BM25)
→ nightly "dream cycle" that enriches pages while you sleep
Hermes memory handles sessions.
GBrain handles institutional knowledge.
2. HERMES WORKSPACE (5.7K stars)
by outsourc-e. Nous Hackathon 2026 winner.
full web-based GUI for Hermes.
chat, terminal, memory browser, skills manager,
agent inspector, live sub-agent streaming.
8 themes. Docker one-click deploy.
what it adds:
→ visual alternative to the built-in dashboard
→ real-time sub-agent activity monitoring
→ embedded terminal inside the workspace
→ file browser with preview (markdown, images, HTML, code)
if you prefer GUI over CLI, start here.
3. MISSION CONTROL (5.3K stars)
by builderz-labs.
multi-agent fleet management dashboard.
dispatch tasks, track costs per agent,
coordinate workflows across profiles.
self-hosted. SQLite-powered.
what it adds:
→ one dashboard for all your Hermes profiles
→ cost tracking per agent per day
→ task dispatch across multiple instances
→ workflow coordination at scale
install this when you run 3 profiles
and need visibility across all of them.
4. SKILLCLAW (1.9k stars)
by AMAP-ML. production-ready.
auto-evolves your skill library
from real session data. adds a post-task
evolution loop on top of Hermes's
built-in skill creation.
what it adds:
→ deduplicates overlapping skills automatically
→ improves skill quality from usage patterns
→ works across multiple devices
→ safety flows: skillclaw doctor hermes
checks health, skillclaw restore hermes recovers
5. AGENTTRACE (v0.5.2)
by luoyuctl. local-first.
TUI for post-run session audits.
reads Hermes logs (plus Claude Code,
Codex, Gemini CLI, OpenClaw, and 7 more).
what it adds:
→ token spike detection across sessions
→ tool failure tracking and retry loop detection
→ latency gaps and health scoring
→ session-to-session diffs and anomalies
→ CI gates (fail builds on unhealthy agent runs)
no upload. no cloud. everything stays local.
brew install luoyuctl/tap/agenttrace
run this after overnight cron jobs
to find what went wrong before it costs more.
which of these are you running?
FULL HERMES NIGHT MODE DIVE 👇