Zeph, an MCP server tool by Bug Ops, is a memory-first AI agent designed for long-running tasks. Built with Rust, it connects to various LLM providers, including local options like Ollama and cloud services such as Claude, OpenAI, and Gemini. Distributed as a single binary, Zeph offers persistent AI assistance, remembering context across sessions for developers, power users, and teams. It manages API keys securely and operates tools within a sandboxed environment.

What does Zeph connect to?

Zeph functions as an MCP server, integrating with a variety of Large Language Model providers. It supports local LLMs like Ollama, alongside cloud-based services such as Claude, OpenAI, and Gemini. Additionally, it can connect to decentralized Trusted Execution Environments. This allows Zeph to route tasks to diverse AI backends, acting as a central point for AI-powered operations. Its architecture is designed to consume services from these LLM providers.

Core features and tools

Zeph is described as a memory-first AI agent capable of long-running jobs. It incorporates sandboxed tools and skill selection, which uses tool-schema filtering for task execution. While the documentation outlines its general capabilities like coding assistance and team chat integration, specific callable tools are not explicitly named. It manages API keys for LLM providers and requires confirmation before executing destructive actions, ensuring secure operations within its sandboxed environment.

Who is it for?

Zeph targets developers, power users, and teams needing continuous AI assistance that retains context across multiple sessions. It is suitable for individuals who require an AI agent to remember project details, code changes, or ongoing research. Teams can deploy it as a bot for chat integration, maintaining conversation history. Skip this if your primary need is a simple, one-off AI query tool without persistent memory or complex task execution.

How to get started

Access is granted through an interactive wizard, "zeph init", which sets up provider keys. These keys are then encrypted and stored in an age vault for secure management of API credentials.

Use cases

• Coding assistance: Point Zeph at a repository to read files, run commands, and search code with full project context.
• Team chat integration: Deploy as a Telegram, Discord, or Slack bot with streaming replies and user allowlists.
• Long-running jobs: Execute research loops, document RAG, scheduled tasks, and multi-step plans.

Zeph provides persistent AI assistance for complex, long-running tasks across development and team environments. Its ability to maintain context and connect to various LLM providers makes it a versatile agent. The secure key management and sandboxed tool execution enhance its operational reliability. Zeph is built in Rust and distributed as a single binary. Try deploying it as a bot in your team chat for persistent AI support.