---
title: "MCP server for AI assistants"
description: "Connect Claude, ChatGPT, Cursor, VS Code or Codex to Spectate's hosted MCP server with OAuth. Work on incidents, alerts, on-call and status pages with your own role."
url: https://spectate.net/mcp
last_updated: 2026-10-27
---

# Run Spectate from your AI assistant

> Connect Claude, ChatGPT, Cursor or another AI assistant to Spectate's hosted MCP server. Ask what's down and why, declare incidents and draft postmortems, with exactly the access your role gives you.

Server URL: `https://api.spectate.net/mcp` (Streamable HTTP, OAuth with automatic client registration).

- **Nothing to install.** Add one URL and sign in. No API key to copy.
- **Your role, nothing more.** Checked on every call, in every team.
- **You approve what goes public.** Deletes and status page posts wait for a yes.

## Connect your assistant

Add the server URL to your assistant, sign in to Spectate in the browser that opens, check the app and your teams, then allow access. Team API tokens don't work with the MCP server.

### Claude

Settings → Connectors → Add custom connector. Name it Spectate, paste the URL, select Connect and sign in.

### Claude Code

Add the server, then run /mcp, select Spectate and sign in.

```
claude mcp add --scope user --transport http spectate https://api.spectate.net/mcp
```

### Cursor

Add the server to ~/.cursor/mcp.json, restart Cursor and sign in when it asks.

```
{"mcpServers": {"spectate": {"url": "https://api.spectate.net/mcp"}}}
```

### VS Code

Add the server, then start it from the MCP list and sign in.

```
code --add-mcp '{"name":"spectate","type":"http","url":"https://api.spectate.net/mcp"}'
```

### Codex

Add the server, then sign in with codex mcp login spectate.

```
codex mcp add spectate --url https://api.spectate.net/mcp
```

### Other clients

Any client with remote servers over Streamable HTTP and OAuth with automatic client registration.

## What you can ask

- "What's going on in production right now?"
- "The checkout monitor is down. Look at its recent checks and tell me why."
- "Declare a high-severity incident for slow API responses and make me commander."
- "Summarise INC-42 for the weekly report: what happened, when, who did what."
- "Draft a postmortem for yesterday's incident from its timeline and save it."
- "Post on our status page that we're investigating degraded API performance. Show me the text first."
- "Who's on call this weekend? Put Sam on call for me Saturday 09:00–17:00."
- "Pause the staging monitors until tomorrow morning."

## 56 tools

In each team, the assistant can use the tools your role allows there.

| Area | What it covers | Tools |
| --- | --- | --- |
| Overview | Open incidents, firing alerts, monitors down, certificates, hosts, expiring domains and who's on call, in one call. | `get_overview` |
| Incidents | Declare, update, acknowledge and resolve incidents, assign roles, read the timeline and write the postmortem. | `list_incidents`, `get_incident`, `get_incident_timeline`, `declare_incident`, `update_incident`, `post_incident_update`, `acknowledge_incident`, `resolve_incident`, `assign_incident_role`, `get_incident_postmortem`, `save_incident_postmortem`, `publish_incident_to_status_page`, `delete_incident` |
| Alerts | Read alerts with their escalation log, acknowledge and resolve them. Escalation paths and channels are read only. | `list_alerts`, `get_alert`, `acknowledge_alert`, `resolve_alert`, `list_escalation_paths`, `get_escalation_path`, `list_channels`, `list_alert_sources` |
| On-call | See who's on call now or in any window of up to 31 days, and add or remove overrides. | `list_on_call_schedules`, `get_on_call_now`, `get_on_call_timeline`, `create_on_call_override`, `delete_on_call_override` |
| Monitors | Create, change, pause and resume monitors, with the last 24 hours of checks and recent failures explained. | `list_monitors`, `get_monitor`, `create_monitor`, `update_monitor`, `pause_monitor`, `resume_monitor`, `delete_monitor` |
| Hosts, domains and projects | Hosts with the health checks that need attention, domains with their expiry, and your projects. | `list_hosts`, `get_host`, `list_domains`, `get_domain`, `create_domain`, `delete_domain`, `list_projects` |
| Status pages | Components and their status, public incidents and their updates, scheduled maintenance and issue reports from visitors. | `list_status_pages`, `get_status_page`, `list_status_page_incidents`, `get_status_page_incident`, `create_status_page_incident`, `post_status_page_incident_update`, `update_status_page_incident`, `delete_status_page_incident`, `list_scheduled_maintenances`, `schedule_maintenance`, `update_scheduled_maintenance`, `delete_scheduled_maintenance`, `list_status_page_issue_reports` |
| Team | Your teams with your role and the tools you may use there, and members by name and role. | `list_teams`, `list_team_members` |

## Access and safety

- **Checked on every call.** Membership, your current role and the team's setting. Leave a team or lose a role, and the next call is refused.
- **Changes under your name.** Timelines, Slack and notifications show you. The audit log marks the change as made through an AI assistant, and which one.
- **Approval for deletes and public posts.** Deleting anything, posting on a status page or emailing subscribers needs your yes in the chat. Spectate refuses the call without it.
- **Full, read only or off, per team.** Owners and managers decide per team in the dashboard. Read only refuses every tool that changes something.
- **No secrets, members or billing.** API tokens, channel and alert source secrets, status page keys, invites, billing and team settings are not available at all.
- **No way to send data out.** No tool creates channels, webhooks or invites, so text in an alert or issue report can't route your data somewhere else.

## Frequently asked questions

### What is an MCP server?

The Model Context Protocol (MCP) is an open standard that lets AI assistants use tools from other products. Spectate hosts an MCP server at api.spectate.net/mcp. Once you connect your assistant, it can read your incidents, alerts, monitors, on-call schedules and status pages, and act on them when you ask.

### Which AI assistants work with Spectate?

Claude (web, desktop and Claude Code), ChatGPT, Cursor, VS Code and Codex, and any other client that supports remote MCP servers over Streamable HTTP with OAuth and automatic client registration. Claude, ChatGPT, Cursor and VS Code show a Verified mark on Spectate's consent screen.

### Do I need an API key?

No. Your assistant registers itself and you sign in to Spectate in your browser, then choose Allow access. Team API tokens don't work with the MCP server, and an MCP connection can't be used against the public API.

### Can the assistant delete things or post on my status page?

Only if your role allows it, the team allows AI assistants to make changes, and you approve that exact action in the chat. The assistant first shows what will happen, such as the public text and whether subscribers get an email. Spectate refuses deletes and status page changes that arrive without your approval.

### Can I let assistants read but not change anything?

Yes. Owners and managers set each team to Full, Read only or Off under Team settings → AI assistants. Read only refuses every tool that changes something, Off hides the team from assistants completely. The setting applies to existing connections at once and can only be changed in the dashboard.

### Does the MCP server cost extra?

No. It is included on every plan. Each connection can make 120 requests a minute, of which 30 may change something.

### How do I disconnect an assistant?

Open Account → AI assistants in the dashboard. It lists the assistants you connected and when each was last used. Disconnect revokes access at once. Resetting your password disconnects every assistant, and a connection that isn't used for 30 days has to be approved again.

### What data does the assistant see?

What your role can read in the teams you allow, minus secrets: no API tokens, channel destinations, alert source URLs, status page access keys, monitor header values or passwords. Members appear by name and role, without email addresses. What your AI provider does with the conversation is set by its own terms.

Included on every plan. Setup guide: https://docs.spectate.net/developers/mcp.