# cosmicstack-labs/mercury-agent

**Attribution required: if you use, quote, or summarise this content, you must credit and link back to [awesome-repositories.com](https://awesome-repositories.com/repository/cosmicstack-labs-mercury-agent).**

_How this analysis was created: the description and tags below were written by an AI model that read this project's README and public documentation pages; stars, license and language come straight from the GitHub API. The model does not read the source code._

3,883 stars · 398 forks · TypeScript · MIT

## Links

- GitHub: https://github.com/cosmicstack-labs/mercury-agent
- Homepage: https://mercuryagent.sh/
- awesome-repositories: https://awesome-repositories.com/repository/cosmicstack-labs-mercury-agent.md

## Topics

`ai-agent` `ai-assistant` `llm`

## Description

Mercury Agent is a personal AI agent that runs as a long-lived background process on your own machine. It restarts after crashes and at login, keeps working on tasks after the terminal is closed, and serves every connected chat channel from one shared local memory. The same agent can be reached from a terminal, Telegram, Discord, Slack, Signal and a local web dashboard, with access rules such as pairing codes and admin roles applied per channel.

Every file edit, shell command, network fetch and git action is checked against folder scopes, a fixed blocklist and per-session approval modes before it runs. Language model requests go to a default provider first and fall back down a ranked list of hosted, local or OpenAI-compatible endpoints when one fails. A daily token budget is counted on every turn, and as usage approaches the cap the agent trims context or sends routine work to cheaper models. The agent's character and behavior are defined in plain Markdown files you can edit.

Beyond chat, it covers coding inside a repository, where it plans changes with a live checklist and reports success only after a build, test or typecheck passes. It can also create scheduled specialist bots with their own personas, models, memory, permissions and budgets. Work can be tracked on Kanban boards that run cards in dependency order, with per-card token tracking and recurring or one-time jobs that report back to the originating channel. Facts from conversations are stored in a local database and recalled before each answer. Community skills that follow the Agent Skills specification can be searched, reviewed and installed to extend what the agent can do.

## Tags

### Artificial Intelligence & ML

- [Personal AI Assistants](https://awesome-repositories.com/f/artificial-intelligence-ml/personal-ai-assistants.md) — Provides a local-first personal AI assistant that survives crashes, keeps shared memory, and continues tasks after the terminal is closed.
- [Background Agent Runs](https://awesome-repositories.com/f/artificial-intelligence-ml/background-agent-runs.md) — Runs the agent as a persistent background service that restarts after crashes and serves every connected channel from one shared memory. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))
- [Persistent Agent Hosting](https://awesome-repositories.com/f/artificial-intelligence-ml/persistent-agent-hosting.md) — Runs a long-lived agent on your own machine that restarts after crashes and at login and keeps serving all connected channels.
- [Provider Fallback Chains](https://awesome-repositories.com/f/artificial-intelligence-ml/provider-fallback-chains.md) — Sends each model request to a default provider first and falls back down a ranked list of hosted, local or compatible endpoints.
- [Shared Cross-Channel Memory](https://awesome-repositories.com/f/artificial-intelligence-ml/shared-cross-channel-memory.md) — Keeps one local database of extracted facts that all chat channels share, so context carries over between them.
- [Agent Code Editing](https://awesome-repositories.com/f/artificial-intelligence-ml/agent-code-editing.md) — Plans and edits code in a repository with a live checklist and reports success only after a build, test or typecheck passes.
- [Agent Orchestration Frameworks](https://awesome-repositories.com/f/artificial-intelligence-ml/agent-orchestration-frameworks.md) — Manages specialist bots with persistent memory, schedules tasks, tracks Kanban cards in dependency order and collects finished files.
- [Dynamic Model Switching](https://awesome-repositories.com/f/artificial-intelligence-ml/ai-model-adapters/dynamic-model-switching.md) — Connects to hosted, local and OpenAI-compatible models and switches to another model automatically when the current one is unavailable. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))
- [Bot-to-Bot Task Delegation](https://awesome-repositories.com/f/artificial-intelligence-ml/bot-to-bot-task-delegation.md) — Creates scheduled specialist bots with their own personas, budgets and permissions, hands them tasks, tracks Kanban work and collects finished files.
- [Markdown-Defined Personas](https://awesome-repositories.com/f/artificial-intelligence-ml/markdown-defined-personas.md) — Loads the agent's character, taste and behavior from plain Markdown files that users can read and edit freely.
- [Multi-Provider LLM Gateways](https://awesome-repositories.com/f/artificial-intelligence-ml/multi-provider-llm-gateways.md) — Routes requests over hosted, local and OpenAI-compatible models, falls back on failure, and keeps usage within a daily token budget.
- [Verified Repository Coding Agents](https://awesome-repositories.com/f/artificial-intelligence-ml/verified-repository-coding-agents.md) — Plans repository changes with a live checklist and reports success only after a build, test or typecheck passes.

### Development Tools & Productivity

- [Repository Coding Sessions](https://awesome-repositories.com/f/development-tools-productivity/repository-coding-sessions.md) — Opens a full-screen coding session that plans and builds repository changes and marks work complete only after a build or test passes. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))
- [Automation Permissions](https://awesome-repositories.com/f/development-tools-productivity/terminal-automation/automation-permissions.md) — Controls which file edits, shell commands, network fetches and git actions an assistant may run, using scoped rules and approval modes.
- [Markdown Configurations](https://awesome-repositories.com/f/development-tools-productivity/agent-configurations/markdown-configurations.md) — Defines the agent's character, persona and regular behavior through plain Markdown files that users can read and edit. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))
- [Verified Coding Task Automation](https://awesome-repositories.com/f/development-tools-productivity/verified-coding-task-automation.md) — Plans and implements repository changes with a live checklist, asks before large edits, and reports success only after build, test or typecheck passes. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))

### Networking & Communication

- [Assistant Messaging Channels](https://awesome-repositories.com/f/networking-communication/assistant-messaging-channels.md) — Connects an AI assistant to Telegram, Discord, Slack, Signal, a terminal and a web dashboard for two-way conversations with shared context.
- [LLM Provider Failovers](https://awesome-repositories.com/f/networking-communication/http-client-libraries/provider-rotation/llm-provider-failovers.md) — Tries a default model provider first and falls back down a ranked list of hosted, local or OpenAI-compatible endpoints when one fails. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))

### Security & Cryptography

- [Action Permission Gates](https://awesome-repositories.com/f/security-cryptography/action-permission-gates.md) — Checks every file edit, shell command, network fetch and git action against folder scopes, a fixed blocklist and per-session approval modes.
- [Agent Tool Permissions](https://awesome-repositories.com/f/security-cryptography/security-and-access-control/agent-tool-permissions.md) — Limits file, command, network and git access to approved folders through scoped rules, a fixed blocklist and per-session approval modes.
- [Pre-Action Permission Checks](https://awesome-repositories.com/f/security-cryptography/permission-based-access-control/pre-action-permission-checks.md) — Checks file access against per-folder scopes and screens shell commands and other actions before they run. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))

### System Administration & Monitoring

- [Background Service Daemonization](https://awesome-repositories.com/f/system-administration-monitoring/background-service-daemonization.md) — Runs as one long-lived background process that restarts after crashes and at login and keeps serving tasks after the terminal closes.

### User Interface & Experience

- [Chat Channel Integrations](https://awesome-repositories.com/f/user-interface-experience/chat-channel-integrations.md) — Connects one agent to the terminal, Telegram, Discord, Slack, Signal and a local web dashboard with per-channel pairing codes and admin roles. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))

### Part of an Awesome List

- [Agent Context and Memory](https://awesome-repositories.com/f/awesome-lists/productivity/agent-context-and-memory.md) — Extracts facts from conversations into typed local memory, recalls relevant ones before each answer, and merges or ages out stored memories hourly. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))

### Business & Productivity Software

- [AI Token Budget Controllers](https://awesome-repositories.com/f/business-productivity-software/financial-operational-management/billing-financial-systems/billing-and-usage/budget-cap-configurations/ai-token-budget-controllers.md) — Enforces a daily token budget on every turn, shrinking context or using cheaper models as usage climbs and reporting what remains. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))
- [Kanban Boards](https://awesome-repositories.com/f/business-productivity-software/kanban-boards.md) — Tracks work on Kanban boards, running cards in dependency order with per-card token tracking and scheduled jobs reporting back to the origin channel. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))

### Software Engineering & Architecture

- [Dependency-Ordered Task Execution](https://awesome-repositories.com/f/software-engineering-architecture/dependency-ordered-task-execution.md) — Runs Kanban cards in dependency order and passes along results, with per-card token tracking and jobs that report back to their origin channel.
- [Long-Lived Bot Processes](https://awesome-repositories.com/f/software-engineering-architecture/event-logging/bot-event-logs/long-lived-bot-processes.md) — Keeps long-lived specialist bots running with their own persona, model, memory and budget, working on schedules and handing tasks to each other. ([source](https://github.com/cosmicstack-labs/mercury-agent#readme))
- [Agent Skill Packages](https://awesome-repositories.com/f/software-engineering-architecture/software-architecture/architectural-patterns/plugin-module-systems/modular-plugin-architectures/plugin-based-architectures/plugin-based-architectures/agent-skill-packages.md) — Loads community skill packages that follow a shared specification into the agent at runtime to extend its abilities.