Open-source LiteLLM plugin

Traffic shaping for LLM spend.

PaceLLM brings traffic shaping to your LLM budget. It throttles low-priority features to cheaper models, protects what matters, and smooths spend across the month.

How it works
Day 25. Everything goes dark.
Every LLM budget tool is one shared credit card. When it hits zero, your $50/hr chatbot dies alongside your $2/hr summarizer.
Without PaceLLM
Day 1$2,000 budget. All features on GPT-4o.
Day 12Engineer starts hammering the internal summarizer.
Day 18Spend rate up 40%. Alert fires. Nobody acts.
Day 25Budget gone. Chatbot down. CEO texts at 11pm.
With PaceLLM
Day 1$2,000 budget. PaceLLM observes traffic.
Day 12Summarizer spike. Auto-downgraded to Haiku.
Day 22Sunday. Idle chatbot budget flows to batch jobs.
Day 30Month ends. All features ran. $82 surplus.
Three things LiteLLM budgets can't do.
โฑ

Pace over time

Learns your traffic. Weekdays get more budget than weekends. You never hit zero on day 25.

๐Ÿ›ก

Protect what matters

Set priorities. When budget tightens, the summarizer downgrades. The chatbot stays on GPT-4o.

๐Ÿ”„

Move idle budget

Chatbot quiet on Sunday? That budget flows to other features. Monday it snaps back.

See where every dollar goes.
Status
On Track
42% spent ยท Day 13 of 30
Spent
$842
of $2,000 this month
Reallocations
3
$18.40 moved today
Forecast
$1,918
projected month-end
Pacing
Use Cases
Forecast
Feb 1 โ€“ 28, 2026

Daily Spend: Planned vs Actual

Planned On track Over 10%

Use Cases

FeaturePriModelSpendStatus
customer_support9gpt-4o$421Normal
product_search6gpt-4o-mini$268Normal
agent_pipeline8gpt-4o$98Strict
internal_summary3haiku$55Downgraded
Plugin is free. Dashboard is the product.
Plugin
$0 forever
Full pacing engine. Open-source. MIT.
  • All pacing & reallocation
  • Unlimited use cases
  • Local dashboard & CLI
  • Webhook alerts
Stop the 11pm text.
Join the waitlist. We'll let you know when it's ready.