For AI Agents

MCP for AI Customer Support

Connect Relvios to AI CLI tools and desktop agents. The MCP server lets trusted agents list tickets, read full customer context, poll new messages, and reply through your autonomous or HITL support workflow.

Available Tools

When you connect Relvios to an AI agent using the Model Context Protocol, the agent can work inside your support inbox with these tools:

list_threads

List customer support tickets. Supports filtering by source, status, priority, and search queries.

read_thread

Read a full thread along with all of its associated messages and internal notes.

send_reply

Directly send a real outbound reply through the configured provider when the inbox is allowed to run autonomously.

list_workspaces

List all workspaces (id and name) where the current API key creator is an admin or owner.

get_new_messages

Fetch new messages across all threads since a specific ISO date string (useful for polling).

Integrations

To install the Relvios MCP server, create a Workspace API key in Relvios. Use per-inbox AI mode settings to decide where agents may act autonomously and where replies must stay human-in-the-loop.

Claude Code

Add the MCP server directly using the CLI command:

$ claude mcp add relvios-uninbox "npx" "-y" "relvios-mcp-bridge"

Gemini CLI

Add the following to your gemini_settings.json:

{ "mcpServers": { "relvios-uninbox": { "command": "npx", "args": ["-y", "relvios-mcp-bridge"], "env": { "RELVIOS_API_KEY": "YOUR_API_KEY", "RELVIOS_BASE_URL": "https://api.relvios.com/api" } } } }

Codex CLI

Add the following to your codex_config.toml:

[mcp_servers.relvios-uninbox] command = "npx" args = ["-y", "relvios-mcp-bridge"] [mcp_servers.relvios-uninbox.env] RELVIOS_API_KEY = "YOUR_API_KEY" RELVIOS_BASE_URL = "https://api.relvios.com/api"

ZeroClaw & OpenCode

ZeroClaw and OpenCode support similar TOML and JSON configurations. Reference the configuration examples in the mcp-configs/ directory of the Relvios codebase for zeroclaw_config.toml and opencode.json.