examples.incidents
Tools: incidents.incidents, incidents.host-status, incidents.crash-report, incidents.runbook, incidents.remediate, incidents.verify, incidents.update-incident. Part of the mocked tool examples.
By ToolGate Examples Signed
Java 17, protocol 2
Use in ToolGate
In the ToolGate console, open Server packages, then Marketplace, and install this package. These values identify exactly what you install:
{
"marketplace": "olo-labs-marketplace",
"packageId": "examples.incidents",
"version": "0.1.0",
"packageDigest": "sha256:8e49bb9f8181693bb961df73655907bd58dbcf857d4a928bcb081612be15ef55"
}
Overview
Server crash recovery (mocked)
Tools: incidents.incidents, incidents.host-status, incidents.crash-report, incidents.runbook, incidents.remediate, incidents.verify, incidents.update-incident. Part of the mocked tool examples.
An on-call assistant for a crashed server, end to end. incidents.incidents lists open incidents. incidents.host-status shows a host's services, CPU, memory, disk and recent deploys. incidents.crash-report gives the crash analysis: the log excerpt, the probable cause and the evidence for it. incidents.runbook returns the steps for that cause. incidents.remediate runs RESTART_SERVICE, ROLLBACK_DEPLOY, FAILOVER or CLEAR_TEMP (once per reference, so a retry never restarts twice). incidents.verify checks the service is healthy again, and incidents.update-incident records the outcome.
Scenario: payments-api on app-02 was killed by the out-of-memory killer after release 2.14.0 (INC-5001, P1); db-01's data disk is filling (INC-5002, P2); web-03 is healthy.
remediate is where a real integration would call your orchestration (Ansible, Rundeck, Kubernetes, a cloud API). Under ToolGate, require approval for it and allow it only for the on-call group.
Try it
cp toolgate.dev.example.json toolgate.dev.json # never commit toolgate.dev.json
toolgate dev # an MCP server over stdio
toolgate pack # build/toolgate/server-recovery-0.1.0.tgpkg
Things to ask a model connected to it:
- "What's on fire right now?"
- "Why did payments-api crash on app-02, and what does the runbook say?"
- "Roll back app-02 to the last good release, check it's healthy, and resolve INC-5001."
- Source repository
- License: Apache-2.0
Tools
incidents.crash-reportRead only- Crash analysis Returns the analysis of the latest crash on a host: probable cause, evidence from the logs, related recent change and confidence.
incidents.host-statusRead only- Host status Returns a host's state, last heartbeat, CPU, memory and disk use, and the state of each service on it.
incidents.incidentsRead only- Open incidents Lists open incidents with priority, host, service and status.
incidents.remediateNon idempotent- Run a remediation Runs a recovery action on a host (mocked: simulated, nothing changes): restart a service, roll back the last deploy, fail over to the standby, or clear temp files and rotated logs. A repeat with the same reference returns the first result instead of acting twice.
incidents.runbookRead only- Get a runbook Returns the recovery runbook for a crash cause.
incidents.update-incidentNon idempotent- Update an incident Sets an incident's status and adds a note to its timeline (mocked: nothing is stored). Stakeholders are notified on MITIGATED and RESOLVED.
incidents.verifyRead only- Verify recovery Runs post-recovery checks for a remediation: service state, health endpoint, error rate and resource use (mocked).
Settings it needs
No settings.
Versions
| Version | API change | Compatibility | Published |
|---|---|---|---|
| 0.1.0 | none | Java 17, protocol 2 | 15 Oct 2026 - 11:55 |
Publisher
ToolGate Examples, OLO Labs
Publisher details are supplied by the publisher. Validation checks package format, not code safety.