Hermes Agent released

Nous Research has open-sourced Hermes Agent, a self-hosted artificial intelligence agent it says can run on a $5 virtual private server and keep working off a laptop. (github.com) The project’s public README says Hermes Agent can run on a $5 server, connect to Telegram, Discord, Slack, WhatsApp and Signal, and use persistent memory, scheduled automations and subagents. Its documentation says the software supports 47 built-in tools and more than 15 messaging platforms. (github.com, hermes-agent.nousresearch.com) Nous Research pushed Hermes Agent v0.9.0 on April 13, 2026 and v0.10.0 on April 16, 2026, according to the project’s GitHub releases page. The April 16 release added a “Tool Gateway” that gives paid Nous Portal subscribers web search, image generation, text-to-speech and browser automation without separate application programming interface keys. (github.com) The trading pitch around Hermes Agent rests on a simple idea: use a language model as an always-on operator that can watch data, size positions and place trades from a cheap cloud machine. Nous Research’s public docs describe Hermes as an autonomous agent that “gets more capable the longer it runs” and can work on cloud infrastructure that “costs nearly nothing when idle.” (hermes-agent.nousresearch.com) One of the claims tied to the demo is Kelly sizing, a bankroll rule that sets how much to bet when the odds appear favorable. George Mason University describes the Kelly criterion as a formula for choosing the percentage of wealth to invest to maximize long-run growth when probabilities are known. (megl.science.gmu.edu) Another claim is weather-market trading, which refers to contracts linked to temperature or other weather outcomes rather than a company’s stock price. CME Group says its weather futures and options are used to hedge weather risk with index-based products tied to seasonal and monthly weather in 18 cities globally. (cmegroup.com) What has not been independently verified in the public project materials is the social-media claim that a Hermes Agent trading setup turned $300 into $123,000 on-chain. The GitHub repository, documentation site and release notes reviewed here describe the agent’s infrastructure and tooling, but they do not provide audited performance records for that trading result. (github.com, hermes-agent.nousresearch.com) That gap leaves two stories running at once. The first is a fast-moving open-source software release with 102,000 GitHub stars and frequent April 2026 updates; the second is a set of trading-performance claims circulating on X that would require separate wallet, exchange or strategy-level verification. (github.com, github.com) Large quantitative trading firms already market themselves around research, engineering and systematic execution. Susquehanna, for example, describes itself as a quantitative trading firm where researchers, engineers and traders design and deploy strategies in a systematic trading environment. (sig.com) Hermes Agent does not turn a $5 server into a prop desk by itself. What it does put in public is a low-cost agent stack that can stay online, call tools, automate workflows and plug into models and messaging systems that used to require much more custom setup. (github.com, hermes-agent.nousresearch.com)