# cosmicstack-labs/mercury-agent

Soul-driven AI agent with permission-hardened tools, token budgets, and multi-channel access. Runs 24/7 from CLI or Telegram.

Repository: https://github.com/cosmicstack-labs/mercury-agent
Canonical: https://ross.abutalabs.com/products/mercury-agent
Homepage: https://mercuryagent.sh/
Language: TypeScript
License: MIT
License Family: permissive
Topics: ai-agent, ai-assistant, llm
Last push: 2026-08-25T07:35:10+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 97, longevity 9
- inputs: {"age_days": 135, "days_push": 8, "days_rel": 23, "gap_med": 0, "n_releases_24m": 30}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3060, forks 329 (observed 2026-08-28T04:07:41.088948+00:00)

## What it is
Mercury Agent is a soul-driven AI assistant that runs 24/7 from the CLI, Telegram, Discord, Slack, Signal, or web, with 40+ permission-hardened tools, a SQLite-backed 'Second Brain' memory, and markdown-defined personality. It enforces token budgets, approval flows for shell/file actions, and supports extensible markdown skills plus multi-agent orchestration and an optional cloud pairing backend.

## Use cases
- run a personal AI assistant from my terminal
- chat with an AI agent over Telegram
- give an LLM agent safe shell and file access with approval prompts
- give my AI assistant persistent long-term memory
- automate daily tasks with an always-on agent
- extend an AI agent with custom markdown skills
- limit my AI agent's daily token spend
- run an AI agent on a Raspberry Pi or Termux

## When to choose
- you want a self-hosted, always-on personal AI agent across CLI and chat channels
- you need permission-gated tool execution with shell blocklists and folder scoping
- you want persistent structured memory with full-text search out of the box
- you prefer a personality defined by editable markdown files you own

## When to avoid
- you need a headless agent framework to embed in your own application rather than a standalone assistant
- you require strict data locality and cannot use the optional cloud backend features
- you need fine-grained multi-tenant agent management at organizational scale

## Facets
- artifact type: application
- maturity: active
- function: agent-framework, llm-inference, chatbot, cli, mcp, rag, scheduling, webhook
- domain: artificial-intelligence, large-language-models, chatbots, developer-tools
- platform: cli, cross-platform, windows, self-hosted
- tags: ai-agent, personal-assistant, memory, token-budget, permission-hardening, skills, multi-agent-orchestration, second-brain, sqlite, tui, telegram-bot, cloud-pairing, ai-agents, command-line, automation, nodejs, macos, linux, telegram, discord, android, web-server

## Member repositories
- cosmicstack-labs/mercury-agent (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:41.088948+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:28:26.223126+00:00, confidence not recorded.
  - readme: https://github.com/cosmicstack-labs/mercury-agent (fetched 2026-08-28T04:07:41.088948+00:00, sha a08a3bbfc0bb)
  - homepage: https://mercuryagent.sh/ (fetched 2026-08-29T09:44:01.776237+00:00, sha dfdfd2722207)
  - site_page: https://mercuryagent.sh/docs (fetched 2026-08-29T09:44:01.785991+00:00, sha ea4a43e4a637)
  - site_page: https://mercuryagent.sh/docs/releases/1.2.0 (fetched 2026-08-29T09:44:01.788439+00:00, sha 9dce78c6694c)
  - site_page: https://mercuryagent.sh/docs/reference/skills (fetched 2026-08-29T09:44:01.790490+00:00, sha 59d638c5ee14)
- Data as of 2026-08-30T08:39:29.467469+00:00.