Token Savings Guide
How Nexus helps you save tokens and reduce costs.
The Problem
Without Nexus, you waste tokens on:
- Re-explaining project structure every session
- Pasting documentation into chat
- Describing conventions repeatedly
- Context that gets lost between conversations
How Nexus Saves Tokens
1. Semantic Search
Instead of pasting entire documentation pages, Nexus returns only relevant chunks:
| Approach | Tokens |
|---|---|
| Paste React docs page | ~5,000 |
| Nexus query result | ~500 |
| Savings | 90% |
2. Persistent Memory
Load all project context with one tool call:
| Approach | Tokens per session |
|---|---|
| Re-explain context | ~2,000 |
| Load from memory | ~200 |
| Savings | 90% |
3. Targeted Results
Nexus returns ranked results with relevance scores, so you get the most useful information first.
Best Practices
Use Specific Queries
Bad: "useEffect"
Good: "useEffect cleanup function for subscriptions"Use Token Formats
json
{
"query": "...",
"tokens": "compact" // or "code-only", "summary"
}Save Once, Use Many
Save project context once at the start, then load it in every session.
Measuring Savings
Track your token usage:
- Note tokens used before Nexus
- Use Nexus for a week
- Compare usage
Most users see 50-90% reduction in context tokens.