mercury-agent vs nanoclaw

Self-hosted AI assistants compared

mercury-agent and nanoclaw are self-hosted TypeScript assistants that connect agents to messaging and scheduled tasks. mercury-agent offers a broad set of interfaces and built-in tools, while nanoclaw emphasizes container isolation and customization through code and skills.

mercury-agentnanoclaw
LanguageTypeScriptTypeScript
LicenseMITMIT
Stars3.2k31.1k
Forks34112.7k
Last analyzedOct 3, 2026Nov 2, 2026

Key differences

  • mercury-agent provides CLI, Telegram, and a local web dashboard; nanoclaw routes messages through supported channels to agent sessions.
  • mercury-agent includes approval controls, folder-level permissions, token budgets, and built-in tools; nanoclaw runs agents in Docker containers with explicitly mounted files and credentials kept outside containers.
  • mercury-agent includes persistent memory, scheduling, and Kanban task boards; nanoclaw focuses on per-agent memory, scheduled tasks, and channel-to-agent arrangements.
  • mercury-agent supports Node.js 20 or later for npm installation and offers standalone installers; nanoclaw expects Node.js 22 or newer and Docker support for containerized agents.
  • mercury-agent has 3.2k stars and 341 forks; nanoclaw has 30.9k stars and 12.8k forks, and its listed latest release is a prerelease.
  • mercury-agent is suited to users configuring a feature-rich assistant; nanoclaw is aimed at technically comfortable users who prefer adapting code over relying on extensive configuration. Both use TypeScript and the MIT license.

Choose mercury-agent if you…

  • need CLI, Telegram, and a local web dashboard for one assistant.
  • want built-in approval controls, token budgets, and a choice of tools including shell, Git, and scheduling.
  • prefer a configurable assistant with persistent memory and Kanban task boards.
Read the mercury-agent analysis →

Choose nanoclaw if you…

  • want agents isolated in Docker containers with explicitly allowed file mounts.
  • prefer customizing the project directly in code and adding installable skills.
  • need separate agent workspaces and memories, or scheduled tasks delivered through messaging channels.
Read the nanoclaw analysis →

Compare the numbers side by side, or add more projects

This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.