Problem solved

Many individuals and small teams want the convenience of an AI assistant but are concerned about data privacy, vendor lock-in, and the limitations of cloud-based bots. They desire an assistant that lives inside their existing chat platforms (like Slack, Discord, Telegram) and can handle context-switching between different projects or personal tasks. Mercury solves this by providing a self-hostable agent that runs on one's own servers, integrates with popular messaging apps, and supports multiple isolated workspaces—all while keeping conversations and data under the user's full control.

Use cases

• Set up a private AI assistant for a team that operates entirely within a Discord server, able to answer questions and execute code snippets.
• Manage separate workspaces for 'Work' and 'Personal' contexts, each with distinct models, triggers, and configurations, accessible via different pairing codes.
• Grant temporary admin access to a colleague by sharing a pairing code, allowing them to manage workspace settings without exposing master credentials.
• Use Mercury as a personal productivity bot that can be summoned in a Telegram chat to summarize meetings, check calendars, or perform quick research.
• Run isolated agent subprocesses with workspace-specific working directories, model selections, and timeout limits for safe experimentation.
• Deploy a tracing-enabled assistant using OpenTelemetry to monitor agent performance and debug multi-step interactions across workspaces.

Audience

• Developers
• Hobbyists
• Startups

Pros

• Self-hosted for full data privacy and control.
• Multi-workspace support enables organizational separation.
• Chat-native integration with messaging platforms.
• Built-in observability via OpenTelemetry tracing.
• Open-source and extensible via skills and plugins.

Cons

• Archived and no longer actively maintained.
• Requires technical expertise to deploy and manage (Docker, CLI).
• Limited documentation and community due to archival.