AlexClaw
A self-hosted AI agent for one user, with scheduled workflows, Telegram and Discord chat, and local models through Ollama or LM Studio.
AlexClaw is a self-hosted personal AI agent for one operator who wants to monitor information sources and automate recurring tasks on their own infrastructure. Built on Elixir/OTP and available through Docker, it collects RSS feeds, web pages, GitHub activity and API data into a searchable knowledge base. It has no multi-user access control.
It supports local models through Ollama and LM Studio, alongside cloud providers such as Gemini and Claude. A task router chooses a configured provider according to the reasoning level the task needs. The agent and its database run on your infrastructure; choosing a cloud provider sends model requests to that service.
Scheduled workflows can fetch information, score its relevance, summarize results and deliver messages through Telegram or Discord. Conditional branches and fallback steps handle different outcomes without asking an LLM to decide the workflow's route. Integrations include Google Calendar, Google Tasks and GitHub diff reviews.
For less predictable work, a reasoning mode breaks a goal into steps, calls permitted skills and evaluates the results. You can pause or steer a run. Persistent memory and a separate reference knowledge base combine semantic search with keyword matching to supply context for later conversations.
Custom skills extend the agent at runtime, subject to declared permissions and code checks. OpenBao stores credentials separately from the database, while two-factor approval protects settings changes and privileged actions. An MCP server lets other clients read stored data and run eligible workflows, but it can't change the agent's configuration.
