avarole-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@avarole-mcp which urgent tickets have nobody assigned to them?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
avarole-mcp
An MCP server for Avarole. It lets Claude, Cursor and other MCP clients look at your Avarole account and work your support queue:
Agents: list them (and the ready-made templates), read their instructions and settings, and (if you allow it) create an agent, or change its instructions, name, voice, greeting, languages and session limit.
Calls: recent calls with length, sentiment, rating and a summary; totals for the last N days.
Support tickets: list and filter them (open, unassigned, by person, by priority, by words), read a whole conversation, see who is holding how many, and (if you allow it) change status, priority, assignee and tags, add private notes, and reply to the customer.
Ask things like "Which urgent tickets have nobody on them?", "Summarise yesterday's calls that went badly" or "Assign ticket 412 to Sue and add a note that I called the customer" or "Create a billing support agent from the support template, greeting callers in English and Spanish".
Set it up
You need an account API key: in Avarole open Developer → API keys → Create key. (Not an agent's embed key, which is meant for websites and is refused here.)
Claude Desktop (claude_desktop_config.json) and Cursor (~/.cursor/mcp.json):
{
"mcpServers": {
"avarole": {
"command": "uvx",
"args": ["avarole-mcp"],
"env": { "AVAROLE_API_KEY": "sk_..." }
}
}
}Claude Code:
claude mcp add avarole --env AVAROLE_API_KEY=sk_... -- uvx avarole-mcp(uvx comes with uv. Or pip install avarole-mcp and use the command avarole-mcp.)
Related MCP server: support-tools
It reads by default
Setting | What it turns on |
nothing | Reading: agents, calls, tickets, workload, statistics. |
| Changing: update a ticket, add a private note, create or change an agent. |
| Emailing customers from a ticket. Off unless you ask for it. |
Add the setting to the env block above. Tools that change things say so in their description, and clients such as
Claude ask you before running them.
Safety
Changing an agent changes how it talks to callers straight away. Avarole silently drops values it cannot use (such as a language it does not speak), so the tool checks what was saved, puts back anything that was refused and says which values did not take.
The key is only sent to Avarole, in a header. It is never put in a URL, a result or a log, and credentials in an agent's settings are never shown.
Ticket and call text comes from customers. The assistant is told to treat it as data. Keep replies off unless you want them, and read what the assistant is about to send.
An account API key can do everything the API allows for your account, so keep it private and revoke it on the Developer page if it leaks.
The tool results go to the AI service you use, so only connect it to accounts whose data you are happy to share with it.
Tools
avarole_list_agents, avarole_get_agent, avarole_list_calls, avarole_call_stats, avarole_list_tickets,
avarole_get_ticket, avarole_ticket_stats, avarole_team_workload, avarole_list_team; with write access
avarole_update_ticket, avarole_add_ticket_note, avarole_create_agent, avarole_update_agent; with replies avarole_reply_to_ticket.
Other settings
AVAROLE_BASE_URL(defaulthttps://app.avarole.com).--transport streamable-httpserves over HTTP on this machine (127.0.0.1) for tools that cannot start a process. It has no sign-in of its own: do not expose it to a network. Hosted use (for example by ChatGPT connectors) would need OAuth, which this version does not include.
Development
pip install -e ".[test]"
pytestApache-2.0. © Avarole.
Available Tools
9 toolsavarole_call_statsARead-only
Totals for the last N days: calls, minutes, average length, sentiment, and the plan's minutes used and left.
Inputs: days.
avarole_get_agentARead-only
One agent's settings: name, voice, instructions (system prompt) and its plain settings. Credentials are never shown.
Inputs: agent_id (required).
avarole_get_ticketCRead-only
One ticket with its conversation (customer messages, replies, private notes) and recent activity.
Inputs: ticket_id (required).
avarole_list_agentsARead-only
List agents (assistants): id, name, voice, and whether it is a template. search matches the name or id.
templates_only lists the ready-made templates a new agent can start from (see avarole_create_agent).
Inputs: limit, search, templates_only.
avarole_list_callsBRead-only
Recent calls and conversations, newest first: when, how long, which agent, sentiment, rating and a summary.
Inputs: limit, offset, agent_id.
avarole_list_teamARead-only
The people tickets can be assigned to (id, name, email).
Inputs: no named parameters.
avarole_list_ticketsARead-only
List support tickets. status: open (default: not resolved or closed), all, or one status. priority: urgent, high, normal or low. assigned_to: unassigned, a person's name, or their user id. search: words in the subject or requester.
Inputs: limit, search, status, priority, +1 more.
avarole_team_workloadARead-only
Who is holding how many live tickets, per person and per group, and how many nobody has taken.
Inputs: no named parameters.
avarole_ticket_statsARead-only
Ticket totals for the last N days: opened, resolved, open now, urgent, average resolution time, by priority and topic.
Inputs: days.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 9 tools
Each tool targets a distinct resource or view: agents, calls, tickets, ticket stats, team workload, and team members. The few potentially similar tools (call_stats vs list_calls, ticket_stats vs list_tickets, team_workload vs list_team) are clearly differentiated by action and output.
All names use the avarole_ prefix and snake_case, with mostly predictable verb_noun patterns such as list_agents, get_agent, list_calls, and get_ticket. The stats and workload tools (call_stats, ticket_stats, team_workload) deviate slightly by using noun-based names rather than verb_noun.
Nine tools is a well-scoped set for an analytics and support overview server. Each tool covers a meaningful slice of the domain without excessive fragmentation.
The surface is largely read-only: agents can be listed and fetched but not created or updated, and tickets can be listed and fetched but not created, replied to, assigned, or resolved. The avarole_list_agents description even references avarole_create_agent, which is not present, creating a clear dead end.
Maintenance
Related MCP Connectors
Hosted support inbox over MCP: answer web chat, WhatsApp, email and SMS, manage knowledge.
Run your SupportWire support inbox from any MCP client: conversations, replies, customers, team.
Remote MCP server for managing Muninx tickets, messages, ticket search, and support analytics.
Work tickets and messages, look up contacts and teams, pull reports, and reply, assign or close.
Related MCP Servers
- FlicenseNot gradedqualityCmaintenanceEnables AI clients to access and manage an internal support ticket queue through MCP tools, resources, and prompts, including searching and viewing tickets, adding comments, closing tickets with confirmation, reading knowledge base articles, and viewing queue summaries over OAuth-secured Streamable HTTP.-
- AlicenseAqualityCmaintenanceProvides customer lookup, transaction history, knowledge-base search, and ticket management tools to any MCP client.5MIT
- AlicenseNot gradedqualityCmaintenanceConnects MCP-capable assistants to support tickets and Agent Tasks through OAuth, exposing read tools for tickets, bots, task runs, and report templates. It also provides scoped write actions such as ticket replies, status and tag updates, draft approvals, task creation, and run decisions.MIT
- FlicenseNot gradedqualityAmaintenanceEnables retrieval and management of tickets and comments, ticket analysis and response drafting, and access to Help Center articles as a knowledge base via MCP.-