HermesAnsem
Everything the agent does
One agent that learns, remembers, reaches every channel, and runs on any model — built to grow more useful the longer you keep it around.
Learns and remembers
HermesAnsem keeps an agent-curated memory: it decides what's worth holding on to and persists it across sessions. It can full-text search every past conversation, summarise what happened, and build a model of you over time — so it stops asking the same questions.
- Autonomous skill creation — after a hard task it saves a reusable skill
- Persistent memory with periodic self-curation
- Searchable history with on-demand summaries
- User modelling that carries between chats
Everywhere you are
A single gateway process runs the same agent — with the same memory — across every place you talk. Start a thread on your phone, finish it in the terminal.
- Telegram, Discord, Slack, WhatsApp, Signal, email and CLI
- Voice-memo transcription in
- One identity and memory shared across all of them
Any model, any host
The agent logic is provider-agnostic. Point it at whatever model fits the task, and deploy it wherever you like.
- OpenAI, Anthropic, OpenRouter, or your own endpoint
- Run local, on a VPS, in Docker, or serverless via Modal / Daytona
- Swap models per task without touching the agent
Tools, automation & subagents
It doesn't just talk — it acts. 40+ built-in tools, MCP servers for anything else, a cron engine for work that runs on its own, and parallel subagents for isolated jobs.
- 40+ integrated tools with an extensible toolset
- MCP (Model Context Protocol) server integration
- Built-in cron scheduler for hands-off automation
- Parallel subagents for isolated workstreams