@kazuph/mcp-devin: AI Agent Orchestration via Slack Conduit

This repository houses a TypeScript implementation of an MCP server specifically engineered to synchronize workflows between the Devin AI platform and Slack team communication channels.

Key functionalities facilitated by this server include:

Initiating autonomous Devin workflows and immediately publishing the resultant tasks as structured messages within Slack.
Enabling seamless message exchange between active Devin processes and their corresponding Slack conversational threads.
Offering sophisticated mechanisms for overseeing and updating the status of these interconnected sessions.

Core Capabilities

Slack Interoperability Layer

Automated task dissemination to Slack, invariably tagging the Devin agent (e.g., using @Devin mentions).
Preservation of conversational continuity by mapping Devin session context directly to Slack threads.
Security and integration achieved via a configured Slack Bot User OAuth Token.

Exposed MCP Toolkit

create_devin_session: Establishes a fresh Devin operational context, posts the initial directive to a specified Slack destination, and returns both the session ID and the Slack message pointer.
send_message_to_session: Submits arbitrary input to a running Devin instance, with an optional parameter to concurrently append the message to the linked Slack thread.
get_devin_session: Retrieves comprehensive metadata for a given session, optionally including the recent conversational history from Slack.
list_devin_sessions: Generates an inventory of all currently managed Devin engagements.
get_organization_info: Fetches configuration details pertaining to the connected Devin organizational entity.

Implementation Guidance

To set up the development environment, execute the following command:

pnpm install

For compilation into deployable JavaScript:

pnpm run build

For development cycles featuring automatic recompilation upon source modification:

pnpm run watch

Configuration Directives

MCP Server Registration

Integration requires adding a specific stanza to your overarching MCP configuration file, mapping the server entry:

"devin-mono": {
  "command": "node",
  "args": ["/path/to/mcp-devin/build/index.js"],
  "env": {
    "DEVIN_API_KEY": "your-devin-api-key",
    "DEVIN_ORG_NAME": "Your Organization",
    "SLACK_BOT_TOKEN": "xoxb-your-slack-bot-token",
    "SLACK_DEFAULT_CHANNEL": "general"
  }
}

Environmental Prerequisites

The following runtime variables must be accurately supplied within the environment block:

DEVIN_API_KEY: Mandatory credential for authenticating with the Devin service.
DEVIN_ORG_NAME: (Optional) Identifies the operating business unit; defaults to "Default Organization".
DEVIN_BASE_URL: (Optional) Specifies the root endpoint for the Devin API; defaults to "https://api.devin.ai/v1".
SLACK_BOT_TOKEN: Essential Slack Bot User OAuth Token (must start with xoxb-).
SLACK_DEFAULT_CHANNEL: The initial channel target for system output. Can be specified by ID (e.g., C123ABC456) or by name (e.g., general).

Deployment Method (Claude Desktop)

To enable this functionality within the Claude Desktop client, update your local configuration file:

On macOS: ~/Library/Application Support/Claude/claude_desktop_config.json On Windows: %APPDATA%/Claude/claude_desktop_config.json

Ensure the mcpServers object includes the following reference:

{
  "mcpServers": {
    "@kazuph/mcp-devin": {
      "command": "/path/to/@kazuph/mcp-devin/build/index.js"
    }
  }
}

Diagnostic Procedures

Because MCP communication utilizes standard input/output streams, debugging can be opaque. We strongly advise leveraging the official MCP Inspector, which is accessible via the package script: