erabot.ai
ProductPricingCase studiesEvalBlogDocs
Book a demo

Swap frontier models for open weights — with proof it won't break.

erabot replays your real traffic through cheaper open-weight candidates, verifies every call site objectively — executed code, structural checks, ground-truth accuracy — adapts the prompt when a swap fails, and hands you signed, expiring proofs. In your environment, under your own keys, so erabot never sees your source.

$ erabot audit .
Book a demo
Runs in your environment under your keys — erabot never sees your code Connect Helicone, Langfuse, or OpenTelemetry and estimates become measured
6/6 task families provably migrated−97% per-call cost on extraction & codegen3/3 failed swaps rescued by adaptation

From the July 2026 validation eval — every number traces to a signed proof artifact.

app.erabot.ai/audit/acme-agent
acme-agentGrade C
audited just now
$4,820
Monthly spend
$1,610
Recoverable
12
Findings
47
Call sites
Recoverable spend by category
Model choice
$1,900
Context size
$980
Caching
$720
Batching
$520
Other
$700
model
api/summarize.py:31
gpt-4o → gpt-4o-mini
SAFE$180/mo
caching
rag/retrieve.py:88
prompt not cached
$95/mo
batching
jobs/backfill.py:12
no batching
$60/mo

No spam. Only launch updates and cost optimization tips.

520
repos in the detection eval
3,049
call sites scored
15
AI products audited
0.80
detection precision

Live Demo

See erabot in action

Watch how we find per-request savings from 25 lines of code.

erabot scan
api_client.py×
config.yaml×
1import openai
2 
3client = openai.OpenAI()
4 
5def analyze_document(doc: str) → dict:
6 # Full document sent as context every call
7 response = client.chat.completions.create(
8 model="gpt-4o",
9 messages=[
10 {"role": "system", "content": SYSTEM_PROMPT},
11 {"role": "user", "content": doc} # 12K tokens avg
12 ],
13 temperature=0.7,
14 )
15 return response.choices[0].message.content
16 
17def summarize_batch(documents: list[str]):
18 results = []
19 for doc in documents:
20 # No caching — identical docs re-analyzed
21 results.append(analyze_document(doc))
22 return results
scan results

How It Works

From codebase to savings in minutes

Four steps. Eight AI agents. Zero setup. Just connect and go.

Step 01

Connect

Run the erabot CLI in your repo, or wire in the GitHub Action. Detects 60 models across 10 providers. Zero config.

Learn more
Step 02

Analysis

Eight AI agents scan in parallel — Token Optimizer, Model Selector, Cache Analyzer, Architecture Reviewer, Cost Projector, Prompt Engineer, Batch Strategist, and Rate Limiter. AST parsing, token counting, and a RAG knowledge base surface every wasted token.

Learn more
Step 03

Optimize

erabot shadow-verifies every fix — it applies the change in-memory and re-runs to confirm quality holds before you see it. You get three output formats: a branded PDF, a Claude Code-ingestible markdown report with code diffs, and git-apply patches.

Learn more
Step 04

Monitor

Connect Helicone, Langfuse, or OpenTelemetry and estimates become measured spend. A CI cost-gate fails a PR when a diff would raise your bill, and realized savings are tracked as fixes land.

Learn more
Your code
Detect
ASTidiomsLLM backstop
Price
Fix + verify
Report + patches

Results

Measured on real codebases

0.80

Detection precision, scored against independent labels on 520 repos

0.73

Full-stack recall — four detection layers, measured on 3,049 labeled call sites

1,086

LLM call sites audited across 15 real open-source AI products

24.7%

Measured cost cut at 100% output quality (escalation routing, replayed traffic)

Model switching, verified — not guessed

The biggest savings come from running a task on a cheaper model — but get it wrong and you ship a quality regression. erabot never recommends a downgrade on a guess. It runs a simulation first:

  1. 1Write representative sample inputs for that exact call site.
  2. 2Replay them through the current model and the cheaper one.
  3. 3A judge scores whether the cheaper model still meets the task.

Only swaps that hold up are shipped. Ones that don’t are shown with their measured quality — never applied automatically.

simulation · title generationreplayed 20 requests
current — gpt-4obaseline
gpt-4o-mini · blanket swap
quality 90% — below the bar
blocked
cheap-first routing
−25% cost · quality 100%
shipped

The unsafe downgrade is blocked and downgraded to a routing fix that keeps quality at 100%.

Output Formats

Three ways to ship fixes

Every scan produces three deliverables — one for each audience.

PDF Report

For stakeholders

Branded executive report with optimization grade, cost breakdown, and per-finding savings. Share with your CTO or finance team.

Download and share — no technical setup needed.

report.md

For humans

Human-readable markdown with rounded savings, per-finding "What to consider" analysis, and no code diffs. Built for skimming and sharing.

Share with your team — human-readable cost analysis.

agent-instructions.md

For AI coding agents

Structured handoff for Claude Code, Cursor, and Copilot. Each finding includes architectural context, fix locus, and anti-patterns to avoid.

Point Claude Code at agent-instructions.md — structured handoff.

Embed erabot in your coding agent

erabot runs where you already work. Drop a single command into Claude Code or Cursor and your coding agent can run the audit and apply the fixes for you — every finding comes as an apply-ready agent-instructions.md it can act on directly.

Download the command file
Claude Code: .claude/commands/erabot.md

Integrations

Works with the tools you already use

Drop erabot.ai into any AI stack — zero migration required.

OpenAIAnthropicGoogle CloudMistralCohereGroqOpenAIAnthropicGoogle CloudMistralCohereGroqOpenAIAnthropicGoogle CloudMistralCohereGroq
LangChainOllamaVercel AIHugging FaceTogetherCursorLangChainOllamaVercel AIHugging FaceTogetherCursorLangChainOllamaVercel AIHugging FaceTogetherCursor

Pricing

Start free. Scale as you save.

No hidden fees. Cancel anytime. Your savings pay for the plan.

Pro

For individual developers serious about costs

$99/mo
  • Unlimited scans
  • 50 files per scan
  • Full MD + PDF reports
  • GitHub repo scanning
  • One-click PR button
  • Email support
Book a demo
Most Popular

Team

For engineering teams optimizing at scale

$499/mo
  • Everything in Pro
  • 200 files per scan
  • 5 seats included
  • Team cost dashboard
  • Shared scan history
  • Priority support
Book a demo

Enterprise

For teams running $10K+/mo on LLM APIs

Custom
  • Founder-led 30-min cost audit
  • Everything in Team
  • 500+ files per scan
  • Unlimited members
  • SSO / SAML
  • Custom DPA
  • SLA guarantee
Book a demo

Not sure which tier? Book a demo

View full pricing comparison →

FAQ

Frequently asked questions

Your next LLM cost reduction starts today.

Book a demo and we’ll run erabot against your codebase live — see your first savings in the call.

EB
erabot.ai

Scan any codebase. See exactly what you spend on AI. Get one-click fixes.

No spam. Only launch updates and cost optimization tips.

Product

  • Product
  • CLI
  • Pricing
  • Changelog
  • Docs

Resources

  • Example audit
  • Blog
  • Case studies
  • Eval

Company

  • Security
  • Privacy
  • Terms

© 2026 erabot.ai. All rights reserved.