SLA & uptime

Service-level objectives for Swiggy MCP endpoints.

Uptime targets

Tier	Monthly uptime SLO	Monthly availability
Production MCP endpoints (`/food`, `/im`, `/dineout`)	99.9%	≤ 43 min downtime/month
OAuth endpoints (`/auth/*`)	99.9%	≤ 43 min downtime/month
Staging endpoints	Best-effort - no SLO in v1	-

Latency targets

Measured at the MCP server edge, excluding client-side network and inference.

Tool class	p50	p95	p99
Read tools (e.g. `search_restaurants`)	`<200ms`	`<600ms`	`<1200ms`
Write tools (e.g. `update_food_cart`)	`<400ms`	`<1000ms`	`<2000ms`
Order placement (e.g. `place_food_order`)	`<800ms`	`<2000ms`	`<4000ms`

Cold-path calls (first tool invocation of a session) may include session validation and can be slower than steady-state.

Status page

A public status page at status.swiggy.com/mcp is shipping in v1.1. Until then, incident information flows via email to your designated engineering contact.

Incident communication

For S0 / S1 incidents, we commit to:

Acknowledgement to your designated engineering contact during business hours (IST). Specific response-time targets are negotiated per partner contract.
Regular written updates until the incident is resolved.
Root-cause analysis shared after the incident is closed.

Maintenance windows

Planned maintenance that may cause brief 5xx responses is announced 72 hours in advance to your designated engineering contact. We avoid windows during Indian Standard Time peak meal hours (12:00-14:00 and 19:00-22:00 IST).

Measurement

SLO is measured as successful request-response pairs / total requests, excluding:

4xx client errors (caller's side)
Requests with revoked or expired credentials
Requests shed by upstream capacity protection (see rate-limits)

Measurement window: rolling calendar month.

Remediation

If SLO is breached in a calendar month, reach builders@swiggy.in for remediation options. In v1 remediation is handled on a partnership basis. Enterprise contracts include explicit credit schedules.