llm-cost-tracker

The first GitHub Action for tracking LLM API costs in CI pipelines.

Teams using AI in CI (code review, test generation, PR summaries) have zero visibility into spend. llm-cost-tracker gives you cost breakdowns, budget alerts, PR comment reports, and artifact output — entirely from built-in pricing tables, no external API required.

Features

• Cost Calculation — built-in pricing tables for OpenAI, Anthropic, Google Gemini, and Cohere
• Budget Alerts — set per-run budget limits, get warned or fail the build on overrun
• PR Comment Report — post a detailed markdown cost breakdown to the PR
• Log File Scanning — scan CI logs for OpenAI/Anthropic response payloads automatically
• Artifact Output — write cost records as JSON or CSV for trend analysis
• GitHub Job Summary — cost table written to the Actions job summary
• No external APIs — all pricing built in, works offline

Quick Start

Direct token input

- name: Track LLM costs
  uses: ollieb89/llm-cost-tracker@v1
  with:
    provider: openai
    model: gpt-4o
    input-tokens: ${{ steps.ai-review.outputs.input-tokens }}
    output-tokens: ${{ steps.ai-review.outputs.output-tokens }}
    budget-limit: '0.50'
    on-budget-exceeded: warn

Log file scanning

- name: Track LLM costs from logs
  uses: ollieb89/llm-cost-tracker@v1
  with:
    log-file: ci-output.log
    budget-limit: '1.00'
    on-budget-exceeded: fail

Full example with PR comment

- name: Track LLM costs
  uses: ollieb89/llm-cost-tracker@v1
  with:
    provider: anthropic
    model: claude-sonnet-4
    input-tokens: '15000'
    output-tokens: '4000'
    budget-limit: '0.50'
    on-budget-exceeded: warn
    post-comment: true
    output-file: cost-report.json
    github-token: ${{ secrets.GITHUB_TOKEN }}

Inputs

Input	Description	Default
provider	LLM provider: openai, anthropic, google, cohere	—
model	Model name (see pricing table below)	—
input-tokens	Number of input/prompt tokens	—
output-tokens	Number of output/completion tokens	—
budget-limit	Budget limit in USD. Triggers alert if exceeded.	—
on-budget-exceeded	warn, fail, or comment	warn
post-comment	Post cost report as PR comment	false
log-file	Path to CI log to scan for LLM usage	—
output-file	Path to write JSON or CSV cost artifact	—
github-token	GitHub token (required for PR comments)	${{ github.token }}

Outputs

Output	Description
cost-usd	Total cost in USD for this run
total-tokens	Total tokens consumed
budget-remaining	Remaining budget in USD
budget-exceeded	true / false

Pricing Table

OpenAI

Model	Input (per 1M tokens)	Output (per 1M tokens)
gpt-4o	$2.50	$10.00
gpt-4o-mini	$0.15	$0.60
o1	$15.00	$60.00
o1-mini	$3.00	$12.00
gpt-4-turbo	$10.00	$30.00
gpt-3.5-turbo	$0.50	$1.50

Anthropic

Model	Input (per 1M tokens)	Output (per 1M tokens)
claude-opus-4	$15.00	$75.00
claude-sonnet-4	$3.00	$15.00
claude-haiku-4	$0.80	$4.00
claude-3-opus	$15.00	$75.00
claude-3-sonnet	$3.00	$15.00
claude-3-haiku	$0.25	$1.25

Google

Model	Input (per 1M tokens)	Output (per 1M tokens)
gemini-2.5-pro	$1.25	$10.00
gemini-2.5-flash	$0.075	$0.30
gemini-1.5-pro	$1.25	$5.00
gemini-1.5-flash	$0.075	$0.30

Cohere

Model	Input (per 1M tokens)	Output (per 1M tokens)
command-r-plus	$2.50	$10.00
command-r	$0.15	$0.60
command	$1.00	$2.00

Aliases supported: gpt4o, sonnet, haiku, opus, gemini-flash, etc.

Example PR Comment

## 💰 LLM Cost Tracker Report

| Metric | Value |
|--------|-------|
| Total Cost | $0.0675 |
| Input Tokens | 15,000 |
| Output Tokens | 4,000 |
| Total Tokens | 19,000 |
| Budget Limit | $0.50 |
| Budget Remaining | $0.4325 |
| Status | 🟢 OK |

### 🤖 Model Breakdown

| Provider | Model | Input Tokens | Output Tokens | Cost |
|----------|-------|-------------|--------------|------|
| anthropic | claude-sonnet-4 | 15,000 | 4,000 | $0.0675 |

Combine with other actions

- uses: ollieb89/ai-code-reviewer@v1
  id: review
  with:
    github-token: ${{ secrets.GITHUB_TOKEN }}

- uses: ollieb89/llm-cost-tracker@v1
  with:
    provider: openai
    model: gpt-4o
    input-tokens: ${{ steps.review.outputs.input-tokens }}
    output-tokens: ${{ steps.review.outputs.output-tokens }}
    budget-limit: '0.10'
    on-budget-exceeded: warn

Tests

npm test

54 tests covering pricing calculations, budget logic, log parsing, report formatting, and CSV output.

License

MIT — see LICENSE

---

Part of the AI DevOps Actions suite

This action is one of five tools that form the AI DevOps Actions suite — the CI/CD layer for AI-native development.

Action	Purpose
ai-pr-guardian	Gate low-quality and AI-generated PRs
llm-cost-tracker	Track AI API costs in CI, alert on overruns
mcp-server-tester	Validate MCP servers: health, compliance, discovery
actions-lockfile-generator	Pin Actions to SHA, prevent supply chain attacks
agent-skill-validator	Lint and validate agent skill repos

→ View the full suite and pipeline example