Provenance Record

News-category (~1,088 words). Overview / What We Know (4 subsections: what it is / Tenacity / Foundation / self-improvement) / What We Don't Know / Analysis structure. Technical and precise. Strong 'What We Don't Know' section: explicitly flags that NVIDIA's OpenRouter-ranking claim is unverified by independent benchmarks, that the OpenAI-compatible local-proxy capability may draw a response from Anthropic/OpenAI/xAI on API-terms grounds, and that OpenRouter daily rankings can swing significantly. Analysis section is genuine editorial work — names the durability gap closed by v0.13 and the distribution gap closed by v0.14.

{"source-0":{"url":"https://github.com/NousResearch/hermes-agent/releases/tag/v2026.5.16","snapshot":"source-0.html.gz","status":"Primary publication (Rule 9): v0.14.0 Foundation release notes. Every PyPI / proxy / Grok / Windows / 180x / 19s / 22-platform specific verified verbatim."},"source-1":{"url":"https://github.com/NousResearch/hermes-agent/releases/tag/v2026.5.7","snapshot":"source-1.html.gz","status":"Primary publication (Rule 9): v0.13.0 Tenacity release notes. Every commit/PR/contributor count + 'finishes what it starts' framing + Kanban/goal/Google-Chat-20th + 8-P0/CVSS 8.1 security wave verified verbatim."},"source-2":{"url":"https://blogs.nvidia.com/blog/rtx-ai-garage-hermes-agent-dgx-spark/","snapshot":"source-2.html.gz","status":"NVIDIA Developer Blog. Verifies 'most used agent in the world according to OpenRouter' framing, '140,000 GitHub stars in under three months' specific, 'Hermes writes and refines its own skills' quote, DGX Spark / RTX framing."},"source-3":{"url":"https://github.com/NousResearch/hermes-agent","snapshot":"source-3.html.gz","status":"Project README. Verifies 200+ models, MIT License, seven backends (local/Docker/SSH/Singularity/Modal/Daytona/Vercel)."},"source-4":{"url":"https://huggingface.co/docs/inference-providers/integrations/hermes-agent","snapshot":"source-4.html.gz","status":"Hugging Face Inference Providers docs. Verifies 'open-source AI agent CLI by Nous Research for coding, research, and development tasks in the terminal' verbatim and 17+ provider framing."}}

Exemplary developer-tools release coverage. Every specific verified verbatim across all 5 snapshots: v0.14 808 commits + 633 PRs + 215 contributors + 165,061 insertions + pip install hermes-agent + hermes proxy command + 'Codex CLI, Aider, Cline, Continue' list + grok-4.3 1M context + native Windows beta + 180x browser-console speedup + 19s cold-start improvement + LINE/SimpleX Chat additions + 22-platform total (v0.14 release notes); v0.13 864 commits + 588 PRs + 295 contributors + 'finishes what it starts' + multi-agent Kanban + /goal primitive + Google Chat as 20th platform + eight P0 security wave + Discord CVSS 8.1 fix (v0.13 release notes); 'most used agent in the world according to OpenRouter', '140,000 GitHub stars in under three months', 'Hermes writes and refines its own skills' quote verbatim, DGX Spark framing (NVIDIA blog); 200+ models, MIT License, seven backends (local/Docker/SSH/Singularity/Modal/Daytona/Vercel) (project README); 'open-source AI agent CLI by Nous Research for coding, research, and development tasks in the terminal' and 17+ providers (Hugging Face docs).