EB
erabot.ai
  • Features
  • How it Works
  • Pricing
  • Blog
Back to Overview
Analysis

See every token your code spends.

Multi-agent AST scanning detects every LLM call, maps costs to call sites, and categorizes findings by impact.

5 Agents. One Scan.

Multi-agent analysis pipeline

Each agent runs in sequence, passing structured findings to the next stage.

CODE INPUTpaste/upload/gitSCANNERAST + patterns5 AGENTSFile · Detect · Cost · Match · BuildSYNTHESIZERmerge findingsREPORTMD + PDF + diff

AST-level call detection

Every LLM call is detected structurally — not with regex. Finds calls in strings, dynamic arguments, and deeply nested code.

Detected:1 LLM call found
summarizer.pypython
import anthropic
 
client = anthropic.Anthropic()
 
def summarize_document(doc: str) -> str:
    response = client.messages.create(
        model="claude-3-5-sonnet-20241022"
        max_tokens=4096,
        messages=[{"role": "user", "content": doc}]
    )
    return response.content[0].text
Findings (3)
HIGH
No context caching
Repeated calls with same context — $0.24/call wasted. Enable prompt caching to cut costs 90%.
MED
Oversized max_tokens
max_tokens=4096 on a task averaging 200 output tokens — billing for 20x unnecessary tokens.
HINT
Model downgrade opportunity
Switch to claude-3-haiku for this task type — save 95% with identical output quality.

4 finding categories

Every finding is tagged by impact type, so your team can prioritize fixes that matter most.

Cost
Token waste, model oversizing, missing cache
e.g. “claude-3-5-sonnet for classification — 50x cheaper alternative exists”
Latency
Blocking chains, no streaming, sync calls
e.g. “3 sequential API calls that could parallelize”
Reliability
No retry logic, missing error handling
e.g. “No exponential backoff on rate limit errors”
Scalability
Per-user API keys, no pooling
e.g. “Creating new client per request — 40ms overhead”

Real cost impact

See exactly how much you're wasting — and what you'll save after applying erabot's recommendations.

Current monthly cost
$347
Based on 10K requests/day
After erabot fixes
$89
74% reduction applied
Save $258/month — with one scan

Most teams see ROI within the first 24 hours of deploying recommendations.

Ready to see it in action?

Scan your codebase in under 60 seconds. No credit card required.