#Copilot CLI Caching Investigation

#Status: No Explicit Cache Support Available

Date: 2026-05-21 Author: Amy & Fry (SquadScope Squad)

#Current Status

#Copilot CLI (Primary Path)

• Copilot CLI in CI mode does NOT support explicit prompt caching
• No flags or configuration options exist to enable cache reuse between runs
• Each invocation sends the full prompt context to the model
• GitHub has not announced any roadmap for prompt caching in the CLI

#GitHub Models API (Fallback Path)

• The GitHub Models API also does NOT expose cache control headers or parameters
• Token usage is billed per-request with no discount for repeated prefixes
• No mechanism to pin or reuse cached prompt segments

#Anthropic API (Direct Access)

• Anthropic's API does support prompt caching via cache_control blocks
• Cached input tokens are billed at $0.30/1M instead of $3.00/1M (for Claude Sonnet 4)
• Cache has a 5-minute TTL with automatic extension on cache hits
• Requires direct API access (not available through Copilot CLI or GitHub Models)

#Cost Impact Analysis

For our current usage (~3 runs/week), potential monthly savings: ~$2.16

#Recommendation

1. Short-term: Continue using Copilot CLI as primary path. The convenience and integration benefits outweigh the caching cost savings at our current volume.

2. Medium-term: Monitor GitHub's announcements for:

   • Prompt caching support in Copilot CLI
   • Cache-aware billing in GitHub Models API
   • Any new --cache or --session flags

3. Long-term / High-volume: If SquadScope scales to daily runs or multi-org deployments, consider switching to direct Anthropic API access to leverage prompt caching. This becomes worthwhile when:

   • Monthly token volume exceeds 1M input tokens
   • The same system prompt is reused across multiple runs within 5 minutes
   • Cost savings justify the added complexity of API key management

#References

• Anthropic Prompt Caching Docs
• GitHub Copilot CLI Documentation
• GitHub Models API