Remote MCP server

Connect AI assistants like Claude, ChatGPT, Cursor, and coding agents directly to Spike over the Model Context Protocol (MCP).

The Spike MCP server lets you connect any AI assistant that supports the Model Context Protocol directly to your Spike data. Query incidents, analyse alerts, check who's on call, acknowledge pages, and manage your on-call configuration — all from the AI tools you already use.

It's a hosted, remote server available at https://mcp.spike.sh/mcp. There's nothing to install and nothing to run — point your assistant at the URL and authenticate.

The server is a stateless bridge to the Spike API. It stores no data and keeps no credentials — every request forwards your own key or token to Spike and reshapes the response for the assistant. It can create and update, but it cannot delete, archive, or restore anything. Irreversible actions stay in the dashboard by design.

What you can do

Incident analysis and reporting

Ask questions across your incident history without manual data gathering:

"How many incidents did we have this week, and what's our MTTR?"
"What are the top offenders this month by incident count?"
"Show me every open sev1 and who owns it right now."

The incident_stats tool returns totals, MTTA/MTTR, and top offenders over any window. Follow the stats → list → show pattern to go from the shape of the data down to a single incident.

Incident management

Create and update incidents, and keep the timeline moving:

"Create a sev2 incident for the payments API degradation."
"Acknowledge the checkout incident and add a note that I'm investigating."
"Resolve INC-456 and set it to p3."

Tools cover create, acknowledge, resolve, unacknowledge, escalate, set priority/severity, reassign, add or remove responders, add notes, and mute/unmute.

On-call and escalation management

Check schedules, see who's carrying the pager, and set up rotations:

"Who's on call for the platform team right now, and who's up next?"
"Who would get paged if payments goes down?"
"Set up a day/night on-call: Ada & Ben 9am–9pm, Cy & Dana overnight."
"Override tomorrow's morning shift with me."

The schedule and layer tools take a per-layer restriction (a daily window or weekly day/time ranges), so day/night and weekday-only rotations are a single request.

Alert routing and outbound webhooks

Inspect and tune how alerts turn into incidents, and where Spike forwards events:

"List our alert routing rules and what each one matches."
"Add a routing rule that sends database alerts to the platform escalation policy."
"Test the PagerDuty outbound webhook."

Configuration and setup

Manage the building blocks of your Spike account — services, escalation policies, integrations, and schedules:

"Create a service for the billing team with a new escalation policy."
"Add a Datadog integration and route it to the payments service."

Team and organisation visibility

See what exists and how it's wired together:

"List all teams and their on-call schedules."
"What escalation policies and integrations does the platform team own?"

Setup

The remote MCP server is available at https://mcp.spike.sh/mcp. There are two ways to authenticate, depending on your use case.

Find MCP in your dashboard

Go to Integrations in the Spike dashboard.
Click New.
Search for MCP — or hit Cmd + K anywhere in the dashboard and search for MCP.

Or jump straight there: https://app.spike.sh/integrations/apps/mcp

For interactive use (OAuth)

If you're a human connecting through an AI assistant like Claude, ChatGPT, or Cursor, authenticate with your Spike account via OAuth. The MCP client handles the flow — you just approve access in your browser when prompted. Actions are attributed to your user account, run with your role's permissions, and show you the same data you'd see in the dashboard.

Spike isn't in the connector directories yet, so add it as a custom connector using the URL below.

Add the remote MCP server:

claude mcp add spike --transport http https://mcp.spike.sh/mcp

Claude Code will prompt you to authorise via your browser on first use.

Open Settings → Connectors.
Click Add custom connector.
Enter https://mcp.spike.sh/mcp and save.
Click Connect and approve access on the Spike consent screen.

Custom connectors are available on Claude's paid plans (Pro, Max, Team, Enterprise). If you don't see Add custom connector, it isn't enabled for your plan.

Add .cursor/mcp.json to your project:

{
  "mcpServers": {
    "spike": {
      "url": "https://mcp.spike.sh/mcp"
    }
  }
}

Cursor opens your browser to authorise on first use.

Open Settings → Connectors.
Click Add custom connector.
Enter https://mcp.spike.sh/mcp and save.
Click Connect and approve access in your browser.

In Settings → Connectors, add a custom connector pointing at https://mcp.spike.sh/mcp and approve access when prompted.

Requires a plan where custom connectors / developer mode are available.

Use the mcp-remote bridge in ~/.codex/config.toml:

[mcp_servers.spike]
command = "npx"
args = ["-y", "mcp-remote", "https://mcp.spike.sh/mcp"]

mcp-remote opens your browser to authorise on first use.

For automated systems (API keys)

If you're connecting programmatic agents, custom workflows, or automation pipelines, use an API key. It authenticates as your organisation rather than a specific user and doesn't expire until you delete it.

Get your API key from Settings → API in the Spike dashboard.
Find your team ID (x-team-id) — the team's _id, visible in the dashboard URL, or returned by the list_teams tool.
Pass both as headers when connecting your MCP client to https://mcp.spike.sh/mcp:

{
  "mcpServers": {
    "spike": {
      "type": "http",
      "url": "https://mcp.spike.sh/mcp",
      "headers": {
        "x-api-key": "YOUR_API_KEY",
        "x-team-id": "YOUR_TEAM_ID"
      }
    }
  }
}

To test connectivity, send the MCP initialize request — every MCP client opens a session with this call, and a successful response confirms the endpoint is reachable, your key is valid, and the protocol version is compatible:

curl -X POST https://mcp.spike.sh/mcp \
  -H "x-api-key: YOUR_API_KEY" \
  -H "x-team-id: YOUR_TEAM_ID" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
      "protocolVersion": "2025-06-18",
      "capabilities": {},
      "clientInfo": { "name": "curl", "version": "0.0.1" }
    }
  }'

x-team-id is required for the incident write tools — without it, the request isn't scoped to your organisation and the server refuses to make it. Eight write tools (reassign, add/remove responders, add note, mute/unmute) are OAuth only and won't appear on an API-key connection, because they must be attributed to a real person.

Getting the most out of it

Read your configuration first

Before filtering, routing, or creating anything, ask the discovery tools for what already exists — list_teams, list_services, list_integrations, and list_escalation_policies give you the IDs and names you'll need for everything else.

Get the full briefing with `get_incident`

get_incident returns an eight-part briefing rather than a raw record — a one-line assessment, the raw alert payload, the execution timeline, who it routes to, blast radius, repeat pattern, and suggested next tools. Reach for it when you want the AI to actually reason about an incident, not just list its fields.

Start with stats, then drill in

For analytical questions, follow the stats → list → show pattern:

incident_stats for the shape (counts, MTTA/MTTR, top offenders)
list_incidents / search_incidents to browse the interesting groups
get_incident for full detail on specific incidents

This is far more efficient than paginating through incidents one by one.

Use the right identifier

Identifiers in Spike are not interchangeable — the single commonest cause of a failed tool call. Each tool's description says where its IDs come from; the rules are:

You want

Use this identifier

x-team-id and any team_id

Team _id from list_teams (not the short slug)

create_incident integration

Integration token from list_integrations

get_integration, update_integration

Integration token (an _id is rejected)

get_service, update_service

Service counterId (a number) from list_services

Escalation policy / alert rule

The entity _id from its list_* tool

remove_override

The override id from list_overrides

Security and privacy

No storage, no secrets. The server keeps no database and no copy of your credentials. Each request forwards your key or token to Spike and forgets it.
Nothing destructive. There are no tools to archive, delete, or restore. The one deletion — removing an on-call override — is reversible by re-creating it. An assistant acting on a bad prompt can't wipe your escalation policies.
OAuth respects your role. Signed-in actions run as you, are limited by your Spike permissions, and are attributed to you in the incident timeline.
Team scoping is enforced. Incident write tools refuse to run without a team ID, so a request is always scoped to your organisation.

Available tools

51 tools are available with an API key; 59 with OAuth. Tools marked OAuth only need a real user identity and appear only on a signed-in connection.

Discovery

Tool

Description

list_teams

List teams and their IDs

list_users

List users in your organisation

list_integrations

List integrations with their tokens

list_integration_types

List available integration types

list_escalation_policies

List escalation policies

list_services

List services

get_integration

Full integration details

get_service

Full service details

get_escalation_policy

Full escalation policy details

Incidents

Tool

Description

list_incidents

Search and browse incidents with filters

get_incident

Full incident briefing — assessment, timeline, routing, blast radius

incident_overview

What needs attention right now

incident_stats

Totals, MTTA/MTTR, and top offenders over a window

get_incident_activity

Activity log for an incident

search_incidents

Search incidents by text

who_owns_incident

Who is on call for this incident this minute

create_incident

Create a new incident

acknowledge_incidents

Acknowledge one or more incidents

resolve_incidents

Resolve one or more incidents

unacknowledge_incidents

Move incidents back to open

escalate_incidents

Escalate to the next level

set_incident_priority

Set priority (p1–p5)

set_incident_severity

Set severity (sev1–sev3)

reassign_incidents (OAuth only)

Reassign to another responder

add_responders (OAuth only)

Add responders to an incident

remove_responders (OAuth only)

Remove responders

add_incident_note (OAuth only)

Add a note to the timeline

mute_incidents (OAuth only)

Mute alerts for incidents

unmute_incidents (OAuth only)

Unmute incidents

On-call

Tool

Description

who_is_on_call

Who is on call now

get_active_shift

The currently active shift for a schedule

list_schedules

List on-call schedules

get_schedule

Schedule details with current and upcoming shifts

who_is_on_call_next

Who is up next

get_user_upcoming_shifts

A user's upcoming shifts

check_user_on_call

Whether a user is on call now

list_overrides

List overrides on a schedule

create_override

Override an on-call slot

remove_override

Remove an override (reversible)

Alert routing

Tool

Description

list_alert_routing_rules

List alert routing rules

get_alert_routing_rule

Full rule details

update_alert_routing_rule

Update a rule

create_alert_routing_rule (OAuth only)

Create a rule

Outbound webhooks

Tool

Description

list_outbound_webhooks

List outbound webhooks

get_outbound_webhook

Full webhook details

update_outbound_webhook

Update a webhook

test_outbound_webhook

Send a test event

create_outbound_webhook (OAuth only)

Create a webhook

Configuration

Tool

Description

create_escalation_policy

Create an escalation policy

update_escalation_policy

Update an escalation policy

create_service

Create a service

update_service

Update a service

create_integration

Create an integration

update_integration

Update an integration

create_oncall_schedule

Create an on-call schedule

update_oncall_schedule

Update a schedule

add_oncall_layer

Add a rotation layer to a schedule

update_oncall_layer

Update a layer

remove_oncall_layer

Remove a layer

Still stuck? Email support@spike.sh.

PreviousIntegrate Spike with Jenkins NextRoles and access

Last updated 23 days ago

What you can do

Incident analysis and reporting

Incident management

On-call and escalation management

Alert routing and outbound webhooks

Configuration and setup

Team and organisation visibility

Setup

Find MCP in your dashboard

For interactive use (OAuth)

For automated systems (API keys)

Getting the most out of it

Read your configuration first

Get the full briefing with get_incident

Start with stats, then drill in

Use the right identifier

Security and privacy

Available tools

Get the full briefing with `get_incident`