Skip to content
LIVCK Cloud

Run LIVCK from your AI assistant

Through the Model Context Protocol, Claude Code, Cursor, VS Code, Codex and many more clients work directly with your organization. 122 tools for incidents, maintenance, monitors and status pages. Anything with an effect on the outside world only happens after you confirm it.

What it looks like on a normal day

Three real flows. You say what you want. The assistant prepares it and asks before anyone hears about it.

You

Checkout has been throwing errors for ten minutes. Open an incident, high severity, and write something decent for our customers.

Assistant

I created an internal draft. The affected service is Checkout, severity major. German and English texts are written. Nobody can see it yet.

Preview

Confirm

Publish on status.example.com

Title: Payment issues at checkout. Status pages: 1. Subscribers to be notified: 418.

Only after you click Confirm does the incident go public and subscribers get their message.

How does LIVCK protect you from assistant mistakes?

A language model can be wrong. That is why a human always makes the final call.

Preview first, then confirmation

Anything that reaches customers, wakes someone, turns alerting off or deletes returns a preview first, with the number of recipients, the text per language and the times.

What you confirm is what goes out

If a teammate edits the text after your preview, LIVCK runs nothing. The assistant has to ask you again with a fresh preview.

No duplicates

If an open incident already covers the same problem, the assistant gets that one back instead of opening a second.

Nothing gets lost

Every edit keeps the previous version, and you can restore it at any time.

Every language is required

If the English version of a customer-facing text is missing, LIVCK rejects the call and says exactly what is missing.

Traceable

Every change made by an assistant shows up in the activity log with the name of its client.

What the assistant deliberately cannot do

See subscriber addresses, read secrets, change members, roles or billing, or change who can access a status page. A connection also never has more rights than the person who created it.

122 tools across 17 areas

This list comes straight from the MCP server's code, so it is always as current as the product.

Basics 2 reads 1 changes
  • confirm_action
  • list_recent_changes
  • whoami
Monitors 4 reads 4 changes 7 with confirmation
  • clear_status_override
  • create_service
  • delete_services
  • duplicate_service
  • get_service
  • inspect_service_payload
  • list_services
  • pause_services
  • preview_service_deletion
  • remove_condition
  • resume_services
  • set_status_override
  • tag_services
  • update_service
  • upsert_condition
Diagnosis 4 reads
  • diagnose_service
  • get_events_timeline
  • get_recent_checks
  • get_service_metrics
Incidents 3 reads 8 changes 8 with confirmation
  • acknowledge_escalation
  • add_incident_comment
  • change_incident_severity
  • create_incident
  • delete_incident
  • delete_incident_message
  • discard_incident_draft
  • edit_incident_comment
  • edit_incident_message
  • find_incidents
  • get_incident
  • page_oncall
  • post_incident_message
  • publish_incident
  • resolve_incidents
  • restore_incident_message
  • suggest_incident_message
  • unpublish_incident
  • update_incident
Maintenance 2 reads 3 changes 9 with confirmation
  • announce_maintenance
  • cancel_maintenance
  • complete_maintenance
  • delete_maintenance
  • delete_maintenance_message
  • edit_maintenance_message
  • find_maintenances
  • get_maintenance
  • post_maintenance_message
  • reschedule_maintenance
  • restore_maintenance_message
  • schedule_maintenance
  • start_maintenance
  • update_maintenance
Postmortems 1 reads 1 changes 2 with confirmation
  • get_postmortem
  • publish_postmortem
  • save_postmortem_draft
  • unpublish_postmortem
Text templates 1 reads 3 changes
  • create_content_template
  • delete_content_template
  • list_content_templates
  • update_content_template
Status pages 2 reads 11 changes 2 with confirmation
  • add_custom_domain
  • create_component
  • create_statuspage
  • create_statuspage_metric
  • delete_statuspage
  • get_statuspage
  • list_statuspages
  • remove_component
  • remove_custom_domain
  • remove_statuspage_metric
  • update_component
  • update_statuspage
  • update_statuspage_layout
  • update_statuspage_metric
  • verify_custom_domain
Subscribers 1 reads
  • get_subscriber_overview
Alerting 3 reads 2 changes 6 with confirmation
  • assign_alert_policy
  • create_escalation_policy
  • create_incident_policy
  • delete_alert_policy
  • get_alert_policy
  • list_alert_policies
  • list_notification_channels
  • send_test_alarm
  • test_notification_channel
  • update_escalation_policy
  • update_incident_policy
On-call 3 reads 4 changes 4 with confirmation
  • create_oncall_schedule
  • create_override
  • get_oncall_now
  • get_oncall_report
  • get_oncall_schedule
  • remove_my_absence
  • remove_override
  • request_cover
  • respond_cover_request
  • set_my_absence
  • update_oncall_schedule
Heartbeats 1 reads 2 changes
  • create_heartbeat
  • list_heartbeats
  • update_heartbeat
Certificates 1 reads 1 changes
  • create_certificate_monitor
  • find_expiring_certificates
Servers 2 reads 1 with confirmation
  • find_servers
  • get_server
  • suppress_server_conditions
Tags 1 reads 3 changes
  • create_tag
  • delete_tag
  • list_tags
  • update_tag
SLA and reports 3 reads 3 changes
  • create_sla_objective
  • delete_sla_objective
  • get_audit_log
  • get_notification_log
  • get_sla
  • update_sla_objective
Support 1 reads 2 changes
  • list_support_tickets
  • open_support_ticket
  • reply_support_ticket

Reads: changes nothing. Changes: changes data, nobody gets a message. With confirmation: reaches customers, wakes someone, turns alerting off or deletes, and only runs after your OK.

Works with your client

The console knows 30 clients and shows you a ready-made configuration with your token for each one.

Terminal

  • Claude Code
  • OpenAI Codex
  • Gemini CLI
  • GitHub Copilot CLI
  • OpenCode
  • Warp
  • Kiro CLI
  • Qwen Code
  • Crush
  • Auggie CLI

Editors and IDEs

  • Cursor
  • VS Code
  • Visual Studio
  • Devin Desktop
  • Kiro
  • JetBrains Junie
  • JetBrains AI Assistant
  • Zed
  • Cline
  • Kilo Code

Desktop apps

  • Claude Desktop
  • Goose
  • LM Studio
  • Jan
  • Msty Studio
  • AnythingLLM
  • Cherry Studio

Self-hosted and automation

  • Open WebUI
  • LibreChat
  • n8n

claude.ai and ChatGPT require sign-in through OAuth. Support for that follows in a later release.

Connected in three steps

  1. 1Create a connectionIn the console under Settings, Organization, AI and MCP. Pick the client, the rights per area, and whether the assistant may notify customers, wake on-call or change alerting.
  2. 2Add the tokenYou see the token exactly once. The dialog drops it straight into your client's configuration, or into an environment variable if you prefer.
  3. 3Start askingAsk your assistant about your monitors or open incidents. Reading works right away, changes ask first.
Claude Code
claude mcp add --transport http --scope user \
  livck https://mcp.livck.cloud \
  --header "Authorization: Bearer lvkm_..."
Cursor · ~/.cursor/mcp.json
{
  "mcpServers": {
    "livck": {
      "url": "https://mcp.livck.cloud",
      "headers": { "Authorization": "Bearer lvkm_..." }
    }
  }
}

What can the MCP server do on each plan?

MCPFreeSoloTeamBusiness
Read
Write–

Reading is included on every plan, the free one too. Writing starts with Solo.

MCP server FAQ

No. Read access is included on every plan, Free included: querying monitors, incidents, response times and failure reasons costs nothing. From Solo at €19 per month excluding VAT, the assistant can also make changes, such as opening incidents, scheduling maintenance or editing status pages.

Claude Code, Cursor, VS Code, Codex and many other clients. The console knows 30 clients and shows the finished configuration for each one, with your token filled in. claude.ai and ChatGPT require an OAuth sign-in, which follows in a later release.

No. Anything that reaches customers, wakes someone, switches alerts off or deletes something first returns a preview with the recipient count, texts per language and times. It only runs after you confirm. If someone changes the text after the preview, the assistant has to ask you again.

Only what the connection is allowed to see, and never more than the person who created it. You grant rights per area. Subscriber addresses, secrets such as passwords and channel tokens, and members, roles and billing stay off limits.

In the console, go to Settings, Organization, AI and MCP, create a connection and pick the client and its rights. You see the token, prefixed lvkm_, exactly once. The server address is https://mcp.livck.cloud. The dialog writes both straight into your client's configuration.

Connect your assistant

Sign up for free, create a connection, ask your first question. Reading costs nothing.

Start for free Read the docs