agent-configs
Control Claude Code, Cursor & Gemini CLI remotely — answer agent questions from your phone via Slack
claude mcp add --transport stdio lassare-hq-agent-configs npx -y lassare-cursor \ --env CURSOR_API_KEY="your-cursor-api-key" \ --env PORTAL_API_KEY="optional-portals-api-key" \ --env SLACK_BOT_TOKEN="your-slack-bot-token"
How to use
Lassare is an MCP server that lets you remotely answer questions and approve actions that your AI coding agents ask for. Each supported agent (Claude Code, Copilot in VS Code, Cursor, and Gemini CLI) has its own MCP configuration folder, enabling a dedicated integration that forwards questions to Slack and returns answers or approvals back to the agent. The system provides two core tools: ask and approve. When the agent needs input, it issues an ask; you receive a Slack DM and respond from your phone. For risky actions, the approve tool provides Approve/Deny options in Slack. Hooks are included to intercept dangerous commands and route them to Slack, helping you stay in control while away from your desk. To use it, sign up, obtain an API key, choose your agent, and follow the setup instructions in the corresponding agent folder README. You can switch between inline and Slack modes with slash commands like /lassare-slack and /lassare-inline. Each agent folder contains its own prompts, hooks, and MCP-specific configuration to tailor behavior and permissions to that agent.
How to install
Prerequisites:
- Node.js installed on your machine (recommended for npx-based setup)
- Access to Slack workspace and bot token
- An Lassare portal API key (for provisioning and management)
Install steps:
-
Clone the repository: git clone https://github.com/lassare-hq/agent-configs.git cd agent-configs
-
Install global or local dependencies if needed (example using npm): npm install
-
Configure environment variables for each agent:
- In the mcp.json or your environment, set SLACK_BOT_TOKEN=your-slack-bot-token CLAUDE_API_KEY=your-claude-api-key CO-PILOT_API_KEY=your-copilot-api-key CURSOR_API_KEY=your-cursor-api-key GEMINI_API_KEY=your-gemini-api-key PORTAL_API_KEY=optional-portals-api-key
-
Start the MCP servers for the agents you enabled. For example:
Claude Code
npx -y lassare-claude-code
Copilot in VS Code
npx -y lassare-copilot
Cursor
npx -y lassare-cursor
Gemini CLI
npx -y lassare-gemini
-
Ensure your MCP.json (or equivalent config) points to these servers and that your agent configurations reference Lassare as the MCP server. When you’re ready, connect your Slack workspace and verify that you can send/receive asks and approvals.
Additional notes
Tips and common issues:
- If a command like rm -rf or git push --force triggers the danger hooks, verify that your prompts and permissions are up-to-date to avoid unnecessary interruptions.
- The hooks and dangerous command list are customizable per agent; adjust permission-approve.sh to fit your security posture.
- If questions or approvals expire after 15 minutes, check your Slack notification delivery and ensure the MCP webhook URL is reachable from your environment.
- Use the two modes (Inline and Slack) to balance speed and safety depending on whether you are at your desk or away.
- Ensure the agent’s API keys and SLACK_BOT_TOKEN are not leaked in logs or version control.
- For troubleshooting, consult the agent folder READMEs (claude-code, copilot, cursor, gemini) for agent-specific prompts, hooks, and setup details.
Related MCP Servers
mcp-local-rag
Local-first RAG server for developers using MCP. Semantic + keyword search for code and technical docs. Fully private, zero setup.
swiftlens
SwiftLens is a Model Context Protocol (MCP) server that provides deep, semantic-level analysis of Swift codebases to any AI models. By integrating directly with Apple's SourceKit-LSP, SwiftLens enables AI models to understand Swift code with compiler-grade accuracy.
sub-agents
Define task-specific AI sub-agents in Markdown for any MCP-compatible tool.
langfuse
A Model Context Protocol (MCP) server for Langfuse, enabling AI agents to query Langfuse trace data for enhanced debugging and observability
mcp_documents_reader
Model Context Protocol (MCP) server exposes tools to read multiple document types including DOCX, PDF, Excel, and TXT. This has been tested on Trae Desktop.
backlog
Help coding agents and developers to keep track of a project's backlog by storing tasks as markdown in git.