Choose theme
AI Resources
Mercury Agent
Mercury Agent is a public Cosmic Stack project for running a personal AI agent from the terminal, a local web dashboard, or messaging channels, with provider choices, tools, skills, persistent memory, scheduled work, and background operation.
Its official materials separate Ask Me from Allow All permission modes, store personality in user-editable Markdown files, add daily token budgeting, and keep Second Brain memory in a local SQLite database. Mercury Code also records whether work finished, paused, was interrupted, or needs recovery instead of treating every stopped run as complete. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A personal agent runtime
Mercury combines a terminal interface, local web dashboard, messaging channels, model-provider choices, tools, skills, memory, scheduling, and background service operation in one public TypeScript project.
Why it stands out
Permissions, memory, budgets, and endings stay visible
The project exposes session permission modes, folder scopes, blocked shell patterns, local structured memory controls, user-owned personality files, token budgets, and completion verdicts instead of hiding those choices behind one assistant interface.
Availability
npm and standalone install paths
The official installation guide lists npm, Bun, pnpm, Yarn, and standalone binary paths for macOS, Linux, Windows, Termux, and lightweight devices. The npm package requires Node.js 20 or newer.
Why it matters
What makes it useful
Mercury is useful when a personal agent has to outlast one terminal session. It can run as a background service, keep its structured memory in a local SQLite database, and use personality Markdown files that the user can edit, giving the reader concrete files and controls to preserve as models or providers change.
What to know
Where it fits
Read Mercury as a configurable personal-agent runtime rather than a finished consumer chatbot. It is most relevant to technically comfortable readers comparing always-on assistants, local memory, provider fallback, messaging access, skills, scheduled work, and permission boundaries.
Notable points
What stands out
Mercury Code v1.2.3 added a completion contract for classifying finished, exhausted, interrupted, truncated, and aborted runs. Its release notes say implementation work must run a build, test, or typecheck before the completion banner appears, while exhausted work can pause and resume instead of being reported as finished.
Before using
What to review
Whether Ask Me or Allow All matches the task. The permissions guide says scheduled tasks run in Allow All mode, so review scheduled work and tool scopes before leaving the agent unattended.
Which model providers, API keys or OAuth accounts, messaging tokens, repositories, folders, shell commands, and external services the setup would reach.
Which personal facts the enabled Second Brain may extract, and how its local SQLite file, pause, search, edit, clear, backup, and file-permission choices should be handled.
How the background service, local dashboard, messaging access, logs, and auto-restart behavior fit the security and maintenance needs of the machine running it.
The version, release notes, platform notes, dependencies, permission defaults, and recovery behavior of the npm package or standalone binary you install before using it with private or important work.
Reader fit
Who may find it relevant
Readers who want a personal agent they can configure, inspect, and keep running on their own machine.
People comparing how agent projects handle memory, provider choice, permissions, skills, budgets, schedules, and messaging access.
Builders who want a public TypeScript codebase plus npm and standalone installation routes.
Less relevant for someone who wants a hosted assistant with no local setup, service management, provider configuration, or permission decisions.
Editorial note
Why LifeHubber lists it
Mercury puts four practical parts of a personal agent in one public project: what it may do, what it remembers, how much token use it can consume, and how it distinguishes finished work from work that needs to resume. That helps readers compare an always-on assistant by its operating boundaries, not convenience alone.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Keep the agent portable after the first setup.
An always-on assistant becomes more useful as it remembers more, which makes its files, permission boundaries, and fallback path worth planning early.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
OpenSEO
every-app/open-seo
An SEO workspace with keyword, ranking, backlink, and site research tools, an MCP connection for AI agents, reusable research skills, and hosted or self-hosted access.
OpenHands Agent Canvas
OpenHands/OpenHands
An OpenHands developer control center for running OpenHands, Claude Code, Codex, Gemini CLI, and custom ACP agents across local, Docker, VM, cloud, and company backends, with conversations, plugins, agent profiles, and scheduled or event-driven automations.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
