New: Replays now compare against the last 90 days.
Read the changelog Status
Capstan status
All systems normal
Every component is working in the US and EU regions. Checked .
Planned: Database upgrade in the EU region
to · EU (Frankfurt)
We're upgrading the EU database. Runs keep going; the app may be slow to load for up to a minute.
Components
Product
App
Workspace, agent builder and run traces
Operational
90 days ago100.00% uptime in the last 90 days, no incidentsToday
API and SDKs
REST API, TypeScript and Python SDKs, CLI
Operational
90 days ago100.00% uptime in the last 90 days, no incidentsToday
MCP server
Capstan tools in Claude, Cursor and other clients
Operational
90 days ago100.00% uptime in the last 90 days, no incidentsToday
Agents
Agent runs
Scheduling and running every step
Operational
90 days ago99.98% uptime in the last 90 days, 1 day with an incidentToday
Approvals
Slack, Microsoft Teams and email
Operational
90 days ago99.99% uptime in the last 90 days, 1 day with an incidentToday
Replays
Testing agents on past cases
Operational
90 days ago99.97% uptime in the last 90 days, 1 day with an incidentToday
Tool connections
Stripe, Zendesk, HubSpot and 120 more
Operational
90 days ago100.00% uptime in the last 90 days, no incidentsToday
Webhooks
Run and approval events to your endpoints
Operational
90 days ago99.99% uptime in the last 90 days, 1 day with an incidentToday
Third parties
AI model providers
The models agent steps call
Operational
90 days ago99.98% uptime in the last 90 days, 1 day with an incidentToday
Email delivery
Sign-in, approval and report emails
Operational
90 days ago100.00% uptime in the last 90 days, no incidentsToday
No incidentsDegraded performanceOutage
Past incidents
October 2026
Slow approvals in Microsoft Teams
Degraded performance · 47 min
Approvals · All regions
Resolved ·
Approval messages in Teams are arriving within seconds again. 212 approvals were delayed by up to 9 minutes; none were lost.
Identified ·
Microsoft's Bot Framework is rate-limiting our requests after a change on their side. We've slowed our retries and are working with Microsoft.
Investigating ·
Some approval requests in Microsoft Teams are arriving late. Slack and email approvals are not affected.
Replays waiting longer than usual to start
Degraded performance · 2 h 10 min
Replays · US (Virginia)
Resolved ·
The replay queue is back to normal. Every queued replay ran; none were dropped.
Monitoring ·
We've added capacity for replays and the queue is shrinking. Live agent runs were never affected.
Investigating ·
Replays in the US region are waiting up to 40 minutes to start, after a spike in 90-day replays.
September 2026
Agent runs failing in the EU region
Outage · 22 min
Agent runs, Replays · EU (Frankfurt)
Resolved ·
Runs are succeeding again. Runs that failed between 07:48 and 08:06 UTC were retried automatically from their last finished step; no step ran twice.
Identified ·
A database failover in Frankfurt left run workers pointing at the old primary. We're moving them to the new one.
Investigating ·
New agent runs in the EU region are failing to start. The US region is not affected.
August 2026
Delayed webhook deliveries
Degraded performance · 35 min
Webhooks · All regions
Resolved ·
Webhook deliveries are back to normal, and every delayed event has been delivered in order.
Investigating ·
Webhook events are being delivered up to 12 minutes late. Runs and approvals themselves are working.
Errors from one AI model provider
Degraded performance · 1 h 5 min
AI model providers, Agent runs · All regions
Resolved ·
The provider has fixed the issue. Steps that failed were retried automatically, and agents with a fallback model never stopped.
Identified ·
One model provider is returning errors for about 20% of requests. Agents with a fallback model are switching to it; others retry with backoff.
Investigating ·
Some agent steps are failing with model errors.