Soul-driven AI agent with permission-hardened tools, token budgets, and multi-channel access. Runs 24/7 from CLI, Telegram or More.
# Mercury Agent
> Mercury is an open-source, soul-driven AI agent that runs on your own machine. It reaches you from the terminal, Telegram, Discord, Slack, Signal and a local web dashboard, keeps a local long-term memory, asks before it acts through a permission system, stays inside a daily token budget, and keeps running as a background service. It ships Mercury Code (a coding environment) and Mercury Bots (persistent specialist agents).
Desktop app: Mercury Desktop for macOS (Apple Silicon and Intel), Windows and Linux at https://mercuryagent.sh/download. It installs Mercury on the computer or connects to one elsewhere, and keeps both updated.
Install: `curl -fsSL https://mercuryagent.sh/install.sh | sh` (macOS/Linux), `irm https://mercuryagent.sh/install.ps1 | iex` (Windows), or `npm i -g @cosmicstack/mercury-agent` on Node.js 22+. Source: https://github.com/cosmicstack-labs/mercury-agent (MIT).
## Docs
- [Getting started](https://mercuryagent.sh/docs): install, first run, channels
- [Mercury Desktop](https://mercuryagent.sh/docs/desktop/mercury-desktop): install the app, first run, connecting to a remote Mercury
- [Updating Mercury](https://mercuryagent.sh/docs/desktop/updating): desktop and server updates, rollback, required updates
- [Remote access & security](https://mercuryagent.sh/docs/desktop/remote-access): owner approvals, device tokens, /server
- [CLI commands](https://mercuryagent.sh/docs/cli-commands/cli-commands): every `mercury` command
- [Permissions model](https://mercuryagent.sh/docs/reference/permissions): scopes, approvals, blocklist
- [Second Brain memory](https://mercuryagent.sh/docs/reference/second-brain): how memory is extracted and recalled
- [Mercury Code](https://mercuryagent.sh/docs/integrations/coding-workspace): the coding environment
- [Mercury Bots](https://mercuryagent.sh/docs/integrations/mercury-bots): persistent bots and fleets
- [Release notes](https://mercuryagent.sh/docs/releases)
## Blog
- [What Is a Personal AI Assistant? 10 Real Tasks in 2026](https://mercuryagent.sh/blog/what-is-a-personal-ai-assistant.md) (2026-11-11): An AI personal assistant acts for one person across their accounts, files, apps and connected services. Here is what it is, ten tasks it can handle, and the permissions each one needs.
- [9 Safest, Most Secure, and Permission-Aware AI Agents for Real-World Tasks](https://mercuryagent.sh/blog/six-ai-agents-compared-for-real-work.md) (2026-11-05): Grok, ChatGPT Work, Mercury, OpenClaw, Manus, Gemini agent for work, Devin, Claude Cowork and Comet compared on memory, where they run, what they can touch, scheduling, permissions, model choice and price.
- [Mercury 1.5: Mercury Desktop](https://mercuryagent.sh/blog/mercury-1-5.md) (2026-11-03): Mercury 1.5.0 ships Mercury Desktop for macOS, Windows and Linux, remote access the owner approves, and updates that roll back on their own.
- [A tour of Mercury Desktop: one app, every Mercury you run](https://mercuryagent.sh/blog/mercury-desktop-tour.md) (2026-11-03): What's inside Mercury Desktop 1.5.0: chat, research, code, bots, memory and voice, plus connecting to several Mercury servers and keeping them all updated.
- [Your agent returned 200 OK. It may still have failed](https://mercuryagent.sh/blog/agent-observability-first-wrong-step.md) (2026-10-31): What to record in an AI agent run to find its first wrong step, and what Mercury records today. Based on OpenTelemetry's GenAI agent spans and a code review.
- [Running an agent overnight: what breaks first](https://mercuryagent.sh/blog/running-an-agent-overnight.md) (2026-10-27): We read Mercury's scheduler and budget code and reproduced one behaviour, to find what fails during long unattended runs: skipped tasks, quiet errors, overlapping jobs and budget stops.
- [How Optimus works: an agent that uses your computer](https://mercuryagent.sh/blog/how-optimus-works.md) (2026-10-24): Inside Optimus, the Mercury plugin that sees your screen, moves the mouse and types: its processes, how it looks and acts, voice control, and the safety rails.
- [Mercury 1.4: plugins, WhatsApp calls, and an agent at your desk](https://mercuryagent.sh/blog/mercury-1-4.md) (2026-10-23): Mercury 1.4 adds isolated plugins and two official ones: WhatsApp chat with voice calls both ways, and Optimus, which sees your screen and uses your mouse.
- [The new bottleneck in AI coding isn't code generation. It's verification.](https://mercuryagent.sh/blog/ai-coding-verification-bottleneck.md) (2026-10-20): A Harvard working paper finds coding agents raised code output while review time rose 49%. Why verification, not generation, should decide whether an agent is useful.
- [Agents exceeded their permissions: how to scope them](https://mercuryagent.sh/blog/agent-permissions-least-privilege.md) (2026-10-17): CERT-In says AI agents read live business records at Indian firms. Here is how least privilege, scoped access, approvals and audit logs should limit them.
- [Mercury Bots: an agent fleet that runs on your machine](https://mercuryagent.sh/blog/mercury-bots.md) (2026-10-16): Always-on agents shipped everywhere this autumn. How Mercury Bots work, what we measured, how they compare with Grok Bot, OpenAI dots, Meta Muse, OpenClaw and Hermes Agent, and the commits behind it.
- [The setup wizard that quit at the first menu](https://mercuryagent.sh/blog/the-wizard-that-quit-at-the-menu.md) (2026-10-15): For six releases, a brand-new Mercury install asked for your name, showed one menu, and exited without an error. A walkthrough of a one-tick race between two readers of the same terminal, and the test that now keeps it from coming back.
- [We were running Mercury twice and didn't know it](https://mercuryagent.sh/blog/we-were-running-mercury-twice.md) (2026-10-14): Typing felt sluggish on every machine, and the terminal renderer took the blame. The real cause was a second full copy of the agent, started quietly by the launcher. How we found it, and why the fix was an architecture change rather than a performance tweak.
- [Mercury 1.3: an agent you can actually leave running](https://mercuryagent.sh/blog/mercury-1-3.md) (2026-10-12): One background runtime that every terminal attaches to, bots that deliver real files and stay online, and a fleet you can run from Telegram. What changed in the 1.3 line and why.
- [Asks before it acts: how Mercury's permission model works](https://mercuryagent.sh/blog/asks-before-it-acts.md) (2026-10-09): An agent with a shell, a file system and network access is only useful if you can trust what it will do with them. A tour of the layers between the model and your machine: scopes, approvals, a blocklist nothing can override, an SSRF guard, and tiers for unattended bots.
## Optional
- [All blog posts, full text](https://mercuryagent.sh/llms-full.txt)
- [RSS feed](https://mercuryagent.sh/blog/rss.xml)