10 Hidden Hermes Tricks Even Veteran Users May Not Know: From /goal to Pets to an API Server
From /goal persistent objectives, hermes import-agent one-command migration, and journey memory starmaps to insights usage analytics, proxy API server, desktop pets, skins, external memory, skill sync, and the iron-proxy egress firewall — 10 Hermes Agent hidden gems, every one verified on a live install.
Read more
Hands-Free Hermes: /heartbeat, /refine, and /goal Gates for Self-Driving Agents
Three new commands merged into Hermes Agent main on 2026-08-06: /heartbeat session heartbeats, /refine on-demand self-improvement reviews, and /goal gate deterministic quality gates. Real CLI examples show how to combine them into a hands-free, self-driving workflow — and where the boundaries with cron and Completion Contracts sit.
Read more
Give Hermes Agent a Semantic Memory with LanceDB: Install, Configure, Benchmark
Hermes' built-in memory matches keywords, so paraphrases miss. LanceDB's official memory plugin adds vector recall + hybrid retrieval: LongMemEval accuracy 0.661 vs 0.533 for the built-in FTS5 baseline. Full install, config, the four memory tools, and a careful reading of the benchmark numbers.
Read more
Stop Paying for Silence: Cloud STT Silence Trim & Local Whisper Idle Unload in Hermes Agent
Voice messages are the most natural way to talk to Hermes Agent — but with cloud STT, every pause in your voice note costs money and invites hallucinated transcript junk. Two new features fix both sides: pre-upload silence trim for cloud providers (measured 13.2s → 6.2s, -53%, identical transcript) and idle unload for the local Whisper model (frees ~370MB of RAM/VRAM, transparently reloads on the next message). This post covers the real config, the honest memory numbers for GPU vs CPU, when to disable the trim, and recommended settings per use case.
Read more
Hermes Desktop Closes the Loop: In-App Browser Lands and the Agent Can Finally Read the Page It Opened
Hermes Agent merged two updates back-to-back: the in-app browser and preview rail became real layout-tree tabs, and a new read_preview tool lets the agent read the visible text of the current page. From open_preview to read_preview, the desktop open-then-read loop is complete.
Read more
Wire Up Hermes Webhooks: Let Your Business Systems Push Alerts to Telegram, Discord, and Slack
Polling wastes API quota and adds latency. Hermes Agent's webhook platform turns order, payment, monitoring, and CI/CD events into instant chat notifications, with Direct Delivery mode for sub-second, zero-token pushes.
Read more
Hermes Smart Approvals: Stop Nodding at Every Step
Approval fatigue is real — every flagged command stops your flow. Hermes v0.19's smart approvals have an LLM review each command, auto-approve the safe ones, auto-deny the dangerous ones, and only escalate the uncertain. Here's how it works and how to keep hard red lines.
Read more
Hermes Agent v0.19.1: The 'Patch' That Touched 4,700 Files and Rebuilt the Voice Stack
v0.19.1 is no ordinary patch. We unpack the official release notes and source commits behind the 4,700-file voice subsystem salvage: full-duplex barge-in, rolling-window VAD, stop-phrase semantics, container repair, and wake-word readiness, with copy-paste config and debugging tips.
Read more
Turn Hermes Conversations into Obsidian Notes Automatically
Stop losing insights inside chat sessions. Wire Hermes Agent's persistent memory, session recall, skills, and cron jobs into an Obsidian vault so your knowledge base updates itself.
Read more