Hermes · the agent
An open-source AI agent that runs on a server you own. It can search the web, run tasks, and chat on Telegram, Discord, Slack, or the command line. You choose the AI model behind it, and you can change it any time.
Three parts working together: your agent does the work, a guard checks the risky moves, and a tamper-proof record keeps track. Swap any part you like. The record stays.
An open-source AI agent that runs on a server you own. It can search the web, run tasks, and chat on Telegram, Discord, Slack, or the command line. You choose the AI model behind it, and you can change it any time.
Before the agent does anything risky, the action is checked. It can be allowed, paused for a person to approve, or blocked outright, and a pause stops the agent until someone decides. You set how strict to be, for each agent.
Every action, and everything the agent learns, is saved to a shared, tamper-proof record. You can search it by meaning, not just keywords, and no one can quietly change what the agent knew or did.
Six views, one password-protected console. Admins control; viewers watch.
Guided setup to a live agent: who it is, its keys, how it behaves. Then a live view to launch, stop, restart, and watch what it's doing.
Every agent in your company in one place: who's live, who went quiet, and what each one has been doing.
Everything the agent remembers, as a daily log. Search it by meaning. Every entry is part of the tamper-proof record.
Actions taken, actions blocked, spend, and active sessions: the numbers that matter, not vanity charts.
Actions paused for a person to approve, the full decision history, and the rules each agent follows.
Step through any past decision: what was done, what was decided, and what got recorded.
Everything your agent learns is saved to a shared, tamper-proof record. It can look things up by meaning, not just exact words, so it remembers what matters, and nothing it knew can be quietly changed later.

See the whole loop run: work done, risk checked, record kept.