BurnRate
Local-first AI coding cost analytics. Tracks Claude Code, Cursor, Codex, Copilot, Windsurf, Cline, and Aider. Cost breakdowns, 23 optimization rules, rate limit monitoring, provider comparison, and PDF reports.
Last verified:
What is BurnRate?
BurnRate is a local-first AI agent observability and cost analytics tool for developers who use AI coding assistants. It tracks costs, tokens, and usage patterns across 7 AI coding providers (Claude Code, Cursor IDE, OpenAI Codex, GitHub Copilot, Windsurf, Cline/Roo Code, and Aider) by reading your existing local session data—no API keys or cloud uploads required. The tool runs entirely on your machine and serves a real-time analytics dashboard at localhost:8089, showing every subagent Claude Code spawns, per-agent costs, tool calls, and session metadata.
Key features include a 46-rule optimization engine that analyzes usage patterns and generates actionable cost-saving recommendations with estimated monthly savings, provider-specific config snippets, and auto-apply capabilities. The dashboard offers 12 views: Overview, Sessions, Costs, Projects, Usage (PRO), Optimize (PRO), Comparison, ROI (PRO), Limits, Budgets, Agents, and Wrapped. BurnRate provides subagent observability (the only tool offering this after Anthropic closed feature requests), real-time rate limit tracking for Claude/Cursor/Codex, per-project budget alerts with desktop/Slack notifications, spend forecasting, and team analytics where only aggregated metrics sync to a shared dashboard—never raw code or session content.
BurnRate is built for individual developers wanting to understand their AI spending without signup or credit card, and for engineering teams needing visibility into AI tool ROI across the organization. The Free tier is completely free forever with no account required, offering 30-day history and basic analytics. Pro ($12/mo) unlocks all 7 providers, unlimited history, the full optimization engine, PDF reports, and ROI analysis. Team ($29/seat/mo) adds team dashboards, per-member cost tracking, team budgets, and Slack/webhook alerts for engineering managers tracking AI spend at scale.
BurnRate pricing
Pricing model: Freemium
Free tier: $0 forever, no account or credit card required. Includes Overview/Sessions/Costs tabs, 7+ providers, 30-day history, basic analytics, rate limit tracking, spend forecasting, and budget alerts (basic). Pro: $12/month, saves $50+/mo on average. Includes everything in Free plus unlimited history, all 12 dashboard tabs, 46-rule optimization engine with auto-apply, PDF report export, provider comparison, model recommendations, operation cost breakdown, context efficiency analysis, ROI analysis, and full budget alerts. Team: $29/seat/month for engineering teams. Includes everything in Pro plus team dashboards, team budget management, Slack/webhook alerts, per-member cost tracking, priority support, member management, and aggregated team analytics. Enterprise: Custom pricing for teams of 50+ with SSO/SAML, dedicated support, custom integrations, SLA guarantee, volume discounts, and on-premise option.
BurnRate pros
- 100% local-first—code and session data never leave your machine
- No signup, no account, no credit card required for Free tier
- Zero config—automatically detects installed AI tools and parses data
- Supports 7 AI coding providers in one dashboard
- Only tool offering subagent observability for Claude Code
- 46 optimization rules with estimated monthly savings per rule
- Provider-specific config snippets with auto-apply to CLAUDE.md
- Real-time rate limit tracking with color-coded utilization bars
- Per-project budget alerts with desktop and Slack notifications
- Spend forecasting with weekday/weekend averages and trend direction
- ROM analysis correlates git commits with AI session timestamps
- Side-by-side provider comparison with efficiency scores (0-100)
- Single binary with zero dependencies—works on macOS, Linux, Windows
- 7-day offline grace period for license validation
- Team analytics sync only aggregated metrics, never raw code
- PDF report export with multi-page executive summary
- Rate limit forecasting predicts when you'll hit usage limits
BurnRate cons
- Cursor IDE has no direct token counts—uses hash count as proxy
- GitHub Copilot has partial token data and no cache metrics
- OpenAI Codex CLI has no cache metrics exposure
- Free tier limited to 30-day session history
- Free tier only includes 3 dashboard tabs (Overview, Sessions, Costs)
- Optimization engine and ROI analysis require Pro paid plan
- Team dashboard requires Team plan at $29/seat/month
- License keys validated against cloud API every 24 hours required for Pro/Team
- Linux ARM64 and Windows require manual download or scoop
- High memory usage possible with thousands of sessions on initial parse
Frequently asked questions about BurnRate
Does my data leave my machine?
Never. BurnRate runs entirely on your local machine. Your code, session data, and analytics are computed locally. The only network call is license key validation for Pro/Team plans, which sends zero usage data.
What is subagent observability?
When Claude Code runs complex tasks, it spawns subagents—child processes that each consume tokens independently. BurnRate parses local JSONL session data and builds a tree view of every agent and subagent: what tools each one called, token usage, duration, and cost. No SDK, proxy, or config needed. Anthropic closed community feature requests (#10164, #10388) for this, so BurnRate is currently the only way to get per-subagent cost breakdowns.
Which AI tools are supported?
Seven providers: Claude Code, Cursor IDE, OpenAI Codex CLI, GitHub Copilot, Windsurf, Cline/Roo Code, and Aider. The plugin architecture makes it easy to add new providers as they emerge. Claude Code provides the richest data with full token counts, cache metrics, and subagent tracking.
How does licensing work?
Free tier works without any account. Pro and Team plans use a license key (64-character hex string prefixed with tt_) entered once via setup wizard, dashboard shield icon, config file, or environment variable. Keys validate against the BurnRate cloud API every 24 hours with a 7-day offline grace period, so it works on planes and in secure environments.
What does the optimization engine do?
It analyzes usage patterns across 46 dimensions: output-heavy sessions, context accumulation, session depth, operation costs, duplicate work, cache efficiency, provider arbitrage, budget overruns, and more. Rules are categorized into Context Efficiency (5 rules), Caching (4), Model Selection (6), Workflow (5), Cost Optimization (4), Tool Usage (3), and Provider-Specific (23). Each recommendation includes plain-English explanations, real-world analogies, estimated monthly savings, and config snippets you can auto-apply.
Can I use this at work?
Yes. Since no data leaves your machine on the Free tier, there are no compliance concerns. Team plans add shared dashboards so your engineering team can track AI tool ROI. Raw code and session data never leave developer machines—only aggregated cost metrics sync to the team dashboard. The architecture is SOC 2 ready by design.
What if I cancel Pro?
You keep the Free tier features forever. Your local data is never affected. You just lose access to the advanced analytics tabs (Usage, Optimize, Comparison, ROI), the optimization engine, PDF reports, and unlimited session history. Your local SQLite cache remains intact.
How does team analytics work?
Each developer installs the BurnRate binary locally and enters their team license key. It analyzes local AI session data and pushes only aggregated cost metrics (total sessions, tokens, estimated cost, model usage, efficiency metrics, project names) to your team dashboard every 15 minutes. Admins see per-developer and per-project spending without ever accessing raw code, session transcripts, or file paths.
Is this SOC 2 compliant?
The architecture is SOC 2 ready by design. User code and session data are processed locally and never transmitted. The cloud component only handles authentication, billing, and aggregated cost metrics. BurnRate is pursuing formal SOC 2 Type II certification.
Do you offer volume discounts?
Yes. Teams of 50+ seats qualify for custom Enterprise pricing with volume discounts, dedicated support, custom integrations, SLA guarantee, and on-premise option. Contact [email protected] for Enterprise pricing.