chronicle:cost-tips — Copilot Billing Is Live, Here's Your Spend Analyzer 💸
Usage-based billing went live on June 1. Two days in, most developers have zero visibility into what's actually burning their premium requests. Chronicle just shipped a new cost-focused command — /chronicle:cost-tips — that surfaces exactly that. Think of it as the spending companion to the workflow-quality /chronicle:tips command we covered in April. It builds on the same session insights infrastructure in VS Code but points the lens squarely at your wallet. This post is commentary on why it matters right now.
Why Cost Visibility Just Became Urgent
Before June 1, Copilot was flat per-seat pricing. There was no incentive to think about how you used it — every request cost the same, with nothing extra. That world is gone. Premium requests now cost real money, per request, and the frontier models are significantly more expensive than standard ones.
The problem: most teams have zero tooling to understand which habits are expensive. Your billing dashboard shows a monthly total, not a behavioral breakdown. Chronicle already stored granular session data locally. cost-tips is just the analytical lens on top of data you were already collecting.
What chronicle:cost-tips Does
It analyzes your local session history for cost-relevant patterns. The key distinction from /chronicle:tips: that command optimizes for workflow quality; /chronicle:cost-tips optimizes for spend reduction.
Recommendations it surfaces:
- Model selection: use standard models for routine tasks instead of defaulting to premium
- Context window bloat: flag sessions where too many files were added to context unnecessarily
- Batching opportunities: suggest combining related small requests instead of many individual turns
- Retry patterns: identify regenerate-spam where refining the prompt would be cheaper
- Outlier sessions: flag sessions consuming disproportionate premium requests relative to output value
Data source is the same local SQLite database — you need github.copilot.chat.localIndex.enabled turned on in your VS Code settings.
Chronicle and all its commands remain experimental. The cost-tips command, its recommendation categories, and the underlying data schema may change or disappear in future VS Code updates.
Enable github.copilot.chat.localIndex.enabled, use Copilot normally for a week, then run /chronicle:cost-tips. The recommendations are only as good as the data — give it at least 5–7 days of real sessions. Focus on model-selection suggestions first; those typically have the highest impact-to-effort ratio.
Practical Scenarios
The accidental premium user. You're using a frontier model for every quick "what does this error mean?" question. Cost-tips flags that 60% of your premium requests were single-turn Q&A that standard models handle identically. Switch those and you cut premium usage by more than half.
The context stuffer. You habitually add entire directories to context when asking about a single function. The extra tokens inflate request cost without improving answer quality. Cost-tips identifies sessions where context size was disproportionate to prompt complexity.
The retry loop. You regenerate 3–4 times hoping for a better answer instead of refining the prompt once. Each regeneration is a fresh premium request. Cost-tips spots the pattern and suggests prompt iteration over brute-force retries.
What It Won't Tell You
- No dollar amounts — it has no access to your billing tier or per-request pricing
- No team aggregation — data is local-only SQLite, not a dashboard for managers
- No coverage of Claude Code or other non-VS-Code agents
- Advisory only — it cannot enforce policies or block expensive requests
- Same schema-stability caveat as all Chronicle features: this may break between VS Code updates
Aliz Stack Connection
Cost management is now a factor in stack decisions. It's not just about which tools are productive — it's about which habits are sustainable at scale.
cost-tips helps individual developers self-optimize before managers have to intervene with blunt policies. That's a better outcome for everyone. It connects directly to our AI Coding Guidelines: model selection guidance is already documented there; cost-tips automates the feedback loop so you don't have to remember the rules.
The gap: if your team uses both Copilot and Claude Code, the Claude sessions remain invisible. Cost-tips only sees what VS Code's session infrastructure captures. Dual-tool teams still need manual awareness for the other half.