Microsoft Copilot · Rate Limits

Microsoft Copilot Rate Limits

Microsoft Copilot rate limits depend on the surface. Microsoft 365 Copilot has per-user fair-use throttling (not publicly published as per-second numbers; soft limits in the high tens of requests per minute per user). Copilot APIs accessed through Microsoft Graph apply Microsoft Graph throttling (per-app, per-tenant, and per-user). Copilot Studio applies per-message metering (counts toward the purchased message pack). GitHub Copilot has per-user request quotas and Copilot Chat conversation pacing. All surfaces use HTTP 429 with Retry-After.

Microsoft Copilot Rate Limits is the machine-readable rate-limit profile for Microsoft Copilot on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 5 rate-limit definitions, measuring requests_per_10s, requests_per_minute, and messages_per_month.

The profile also includes 5 backoff/retry policies defined and response codes documented for throttled and serviceUnavailable.

Tagged areas include Agents, AI Assistant, Artificial Intelligence, Chatbot, and Copilot.

5 Limits Throttle: 429
AgentsAI AssistantArtificial IntelligenceChatbotCopilotExtensibilityGenerative AIRate Limiting

Limits

Microsoft Graph — per-app per-tenant app/tenant
requests_per_10s · second
130000
130,000 requests per 10 seconds per app per tenant for general Graph endpoints. Many service-specific limits override (Outlook, OneDrive, etc.).
Microsoft Graph — per-user user
requests_per_10s · second
10000
10,000 requests per 10 seconds per user across all apps. Soft limit subject to change.
Microsoft 365 Copilot — fair-use per user user
requests_per_minute
see vendor docs (soft fair-use)
Per-user fair-use throttling on Copilot prompts; not publicly documented as per-second numbers. Excessive prompting yields 429 with Retry-After.
Copilot Studio — message metering tenant
messages_per_month · month
25000
Standard message pack is 25,000 messages/month. Excess billed PAYG at $0.01/message. Generative messages count as 2; classifier as 1.
GitHub Copilot — chat requests user
requests_per_minute
see vendor docs
Per-user throttling on Copilot Chat; specific numbers not published. Code completion uses streaming and is tuned per-IDE.

Policies

Honor Retry-After
All Copilot surfaces return 429 with Retry-After. Implement exponential backoff with jitter; SDKs auto-retry.
Microsoft Graph backoff
Microsoft Graph documents service-specific throttling and recommends batching via /v1.0/$batch (max 20 per batch) plus delta queries for large reads.
Per-message accounting
Copilot Studio monthly message packs are not pooled across tenants. Unused messages don't roll over. Excess goes to PAYG.
Tenant fair-use
Sustained extreme load on Microsoft 365 Copilot can trigger tenant-level throttling and slowdown across users. Rare but documented.
Customer-managed key impact
Tenants using Customer Lockbox / CMK may see slightly lower throughput due to additional cryptographic operations.

Sources