Hermes Agent v0.5.0 Supports 400+ Models, Hardens Supply Chain
TL;DR
Hermes Agent v0.5.0 (March 28, 2026) routes 400+ models via Nous Portal, adds Hugging Face as a first-class provider, removes the compromised litellm dependency, and exposes plugin lifecycle hooks.
What changed
Hermes Agent v0.5.0 (The Hardening Release) shipped March 28, 2026. Nous Portal now routes to 400+ models through a single provider endpoint. Hugging Face becomes a first-class inference provider with a setup wizard, live /models endpoint probe, and a curated agentic model picker mapping HF models to OpenRouter analogues. Four plugin lifecycle hooks land in the agent loop and CLI/gateway: pre_llm_call, post_llm_call, on_session_start, on_session_end. Telegram private chat topics support project-based conversations with per-topic skill binding. Supply-chain work removes the compromised litellm dependency, pins all dep version ranges, regenerates uv.lock with hashes, and adds CI scans for supply-chain attack patterns.
Why it matters
The litellm removal directly addresses the 2025 compromise: no external dependency whose security posture you cannot verify sits in the inference path anymore. Plugin lifecycle hooks turn token budgeting, PII redaction, logging, and cost tracking into composable extensions rather than core patches. Nous Portal at 400+ models gives you a managed alternative to OpenRouter through one billing relationship. Per-topic skill binding on Telegram lets you isolate workflows without spinning up multiple bots.
What to watch for
Upgrade is on the standard path; the litellm removal is transparent because existing calls route through Hermes's native provider interfaces. Verify that any custom integrations you previously layered over litellm now resolve through the native adapter. Test the new lifecycle hooks against a non-production session before wiring them to redaction or budgeting in production. Audit your CI to mirror the Hermes supply-chain scan patterns where applicable.
Who this matters for
- Developers: Plugin lifecycle hooks (pre/post LLM, session start/end) let you add PII redaction, token budgeting, or logging without forking. Pinned dep version ranges and hashed uv.lock for reproducible builds. litellm removed in response to the 2025 compromise.
Harsh’s take
The supply-chain hardening is the story most people will skip over. Removing litellm is not a small call: litellm is widely used as the adapter for every model provider, and dropping it means Hermes had to implement its own adapter layer for every provider. The team made that choice deliberately after the 2025 litellm compromise. For anyone running Hermes against production data, this is the risk-posture move that matters most, and the hashed uv.lock plus pinned ranges close the surrounding reproducibility gaps.
Nous Portal at 400+ models is the subscription economics play: one subscription, one integration, full frontier model access through a single endpoint. Plugin lifecycle hooks (pre_llm_call, post_llm_call, on_session_start, on_session_end) are the extensibility foundation the ecosystem needs. Token budgeting, PII redaction, logging, cost tracking: all of these become out-of-tree plugins rather than fork-and-patch jobs. Expect a wave of community plugins exploiting these hooks in the coming weeks.
by Harsh Desai
About Hermes Agent
View the full Hermes Agent page →All Hermes Agent updatesMore AI news
- Daily RoundupVercel Sandbox drives, Gemma 4 QAT, and NVIDIA models trending on Hugging Face (agent tools today)
Vendors added persistent storage and quantized models while new text-to-speech and image models appeared on Hugging Face and Fal, with fresh agent tools listed on Product Hunt.
- Weekly DigestCursor enterprise orgs, Claude Code Bedrock auto mode, Codex Sites preview for quick deploys
Cursor added team management and design tools while Claude Code and Codex rolled out cloud integrations, safety checks, and deployment features across CLI and apps from 29 May to 5 June 2026.
- Daily RoundupGemini for Education hits Utah K-12, Nemotron 3 Ultra on Vercel, and agent workspaces from Product Hunt
Google expanded Gemini access in schools and infrastructure while Nvidia and Vercel enabled heavier agent models, and Product Hunt surfaced new local and collaborative agent tools.