# Triage digest: Hermes Agent local LLM free tier
Generated 2026-10-31. 15 real search results classified into HOT / FOUNDATIONAL / NOISE.

## Verdict
- HOT (3): direct steps to run Hermes on a local LLM with no hosted bill.
- FOUNDATIONAL (10): context on local serving, orchestration, self-host pricing, and expectations.
- NOISE (2): generic vendor price lists, useful for cost math only.

## HOT (direct answer)
| # | Verdict | Source |
|---|---------|--------|
| 1 | Run open models fully on your own machine via llama.cpp; free inference, no hosted token bill. | https://hermes-agent.nousresearch.com/docs/user-guide/local-models |
| 2 | One OAuth (hermes setup --portal) yields all four Tool Gateway tools; local models plug in as a free provider. | https://hermes-agent.nousresearch.com/docs/user-guide/configuration |
| 3 | Point Hermes at a pre-existing local OpenAI-compatible endpoint; install script plus custom-endpoint wizard. | https://build.nvidia.com/playbooks/hermes-agent/instructions |

## FOUNDATIONAL (context)
| # | Verdict | Source |
|---|---------|--------|
| 4 | Hermes memory can be self-hosted/OSS at no license cost; separate from local inference. | https://github.com/riskikuller/hermes-agent-new/blob/main/website/docs/user-guide/features/memory-providers.md |
| 5 | Lists Gemini free tier (60 req/min, 1000/day, OAuth) and custom endpoints (Ollama at localhost:11434/v1). | https://github.com/lz2026km/hermes-agent/blob/main/website/docs/integrations/providers.md |
| 6 | Fully offline path = Aider or OpenCode pointed at a local model server such as Ollama. | https://www.getorchestra.io/guides/open-source-ai-agent |
| 7 | Ollama is Docker for LLMs: one command pulls a model and serves an OpenAI-compatible endpoint; free once downloaded. | https://www.kdnuggets.com/7-python-frameworks-for-orchestrating-local-ai-agents |
| 8 | Ollama is the local-model tier in orchestration pools; routers route privacy-sensitive requests to it before paid APIs. | https://www.ayautomate.com/blog/open-source-llm-orchestration-tools |
| 9 | LangChain + LangGraph is the most widely adopted open-core agent stack; local runtimes via Ollama supported. | https://www.langchain.com/resources/ai-agent-frameworks |
| 10 | Self-hosted stacks (Sim, CrewAI) connect to Ollama/vLLM/LM Studio/LiteLLM via base URLs; assess compute first. | https://www.sim.ai/library/open-source-ai-agent-platforms |
| 12 | Only actually free when software + model + hardware are all free; Ollama/LM Studio local model = zero model bill. | https://agnt.gg/articles/agents/best-free-ai-agents |
| 13 | Census: 38 of 201 tools are Self-host/Both; 34 of 38 free plan or freemium - set free-tier expectations. | https://theagentsindex.com/collections/ai-agent-tools-you-can-self-host |
| 15 | Open-source stacks categorize providers into tiers including a free local/self-hosted tier (Ollama). | https://github.com/DevAnimecx/free-ai-agent-stack/blob/main/data/llm-apis.yaml |

## NOISE (cost math only)
| # | Verdict | Source |
|---|---------|--------|
| 11 | Vendor-priced comparison table; useful for cost math, not for picking this stack. | https://infrabase.ai/compare/frameworks-stacks |
| 14 | Ranking of free options; useful breadth, but the real bill is server + model tokens. | https://resources.rework.com/tools/ai-agents/best-free-ai-agents-2026 |