# atris [Health: Active]

**Category:** 🤖 Coding Agents  
**Repository:** https://github.com/atrislabs/atris  
**GitHub Stars:** 71  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/atris

## Description
Say what you want in plain words. Atris plans it, builds it, checks it, and shows proof. CLI plus MCP server for AI agents.

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What the atris MCP server does

The atris MCP server gives an AI agent an operating workspace around an existing repository. Atris initializes an `atris/` directory containing instructions, navigation, task views, daily logs, feature packs, reusable skills, team context, and wiki memory. Agents can read this workspace before acting and use it as the local source of operating context.

Work follows a `plan -> do -> review` cycle. Atris can handle individual objectives, recurring missions, member-managed jobs, project-memory checks, and recap generation. The system is intended for more than code: the README also describes research, CRM, reporting, event operations, support, and general business operations as possible job types.

## How it works

The atris MCP server operates on a workspace associated with a repository. `atris init` creates the initial structure, and running `atris` loads the workspace context before handing work to the instructions in `atris/atris.md`. The repository also gains files such as `TODO.md`, `MAP.md`, dated logs, feature validation material, and wiki status.

Tasks can include a verification command. Member jobs run in separate repository copies, complete their own check, and then receive review from a different engine. Results include the work performed and a receipt. Failed checks, review failures, unsaved changes, branch conflicts, and protected-area changes can prevent automatic landing. A job that fails three consecutive times pauses for human attention.

The command-line workflow includes `atris run` for one objective, `atris brainstorm` for early ideas, `atris autopilot` for continued missions, and `atris loop` for checking stale or incomplete project memory. `atris recap` produces a summary, with `--share` providing a receipt-backed version.

## Setup and configuration

Install Atris globally with npm, then initialize it from the target project:

```bash
npm install -g atris
cd your-project
atris init && atris
```

Node.js 18 or newer is required. The `atris task` command uses built-in SQLite and requires Node.js 22 or newer. Run `atris setup` when cloud workspaces, businesses, or integrations are needed.

Agents can sign in without an interactive browser by creating an agent client and storing its secret in a private file. The documented environment variables are `ATRIS_CLIENT_ID`, `ATRIS_CLIENT_SECRET_FILE`, and `ATRIS_CLIENT_SECRET`; an existing `ATRIS_TOKEN` takes precedence. The README does not specify an MCP transport or client-specific configuration.

## Tools and capabilities

The atris MCP server is suitable for agents that need:

- Persistent workspace instructions and repository memory
- Durable tasks with explicit verification commands
- Daily logs, feature packs, skills, and team-member context
- Member execution, wake, sleep, result, and tidy workflows
- Review receipts and recap summaries
- Optional self-landing through a branch, pull request, and GitHub auto-merge flow
- Static-site publishing through the Atris CLI

Static sites can be published with `atris site publish`, including a documented profile and build option for a Next.js static export. Member work can be watched through the board or progress logs, and work remains in its copy until it is landed or tidied.

## Limitations and notes

The atris MCP server is documented alongside a larger CLI and workspace system, but the provided material does not enumerate individual MCP tool names, transport options, or supported MCP client applications. Cloud workspaces and integrations require the separate `atris setup` step. Automatic landing is opt-in and does not proceed when checks or reviews fail, when the working tree is unsaved, when branches conflict, or when protected areas are involved. AgentXP acceptance also requires human review to mint Career XP.

_Full upstream README: https://allmcps.com/mcp/atris/readme_