Last updated: 2026-08-19
Carrot Labs launched SuperPenguin through Y Combinator's W26 batch to attribute AI spending that otherwise arrives as one lump provider bill. The platform syncs data from 14 providers and tags each request to a customer, feature, team, or pull request through SDKs that never capture prompt or response content.
About SuperPenguin
Carrot Labs built SuperPenguin, an AI spend intelligence platform, after its founders Christopher Acker (formerly Skylo Technologies) and Yuta Baba (formerly Snowflake) kept hitting the same wall: AI bills arrive as one lump sum per provider, with no way to trace a dollar back to the customer, feature, or engineer that caused it. The company launched through Y Combinator's Winter 2026 batch to fix that gap directly. The platform syncs cost data from 14 AI providers, including OpenAI, Anthropic, Google Gemini, AWS Bedrock, Azure, Deepgram, ElevenLabs, Modal, and OpenRouter, once an API key is connected. It is deliberately not a proxy: requests travel straight to the provider, only cost metadata is captured, and every estimate gets checked against the provider's own invoice rather than trusted on its own. A companion Mac app extends the same idea to engineering spend, pulling GitHub pull request data so a team can see what shipping a feature actually cost in AI usage. The intended buyer is an engineering lead or early-stage founder who currently reconstructs this picture by hand from exported CSVs, not a large enterprise with a dedicated FinOps function. Pricing scales with the AI spend a team actually manages rather than seat count, with a no-card-required entry point and negotiated terms once usage grows past the published self-serve tiers.
Pricing
Free: $0/month, up to $2,000 in managed AI spend, no card required. Growth: $30/month, up to $5,000 managed spend, 3 team members. Pro: $200/month, up to $20,000 managed spend, 10 team members. Enterprise: custom pricing, $20,000+ managed spend, unlimited team members. Billed on spend managed, not per seat or per request.
Key Features
- Automatic provider cost sync: Adding an API key from a supported AI provider, such as OpenAI, Anthropic, or AWS Bedrock, pulls in costs automatically with model-level breakdowns, trends, and forecasts, no code required.
- Per-request attribution SDKs: Python and TypeScript SDKs tag every AI call with a customer, feature, team, or environment label, breaking a provider's monthly total down to the individual request.
- Coding cost per pull request: A free Mac menu bar app links to GitHub to price each pull request across Cursor, Claude Code, and Codex, so engineering leads see what a shipped feature cost in AI usage.
- Configurable spend alerts: Thresholds and anomaly detection trigger notifications to Slack, Discord, or email when spend spikes, catching a runaway agent loop before it reaches the invoice.
- Invoice reconciliation: SDK cost estimates are checked against the provider's actual billed invoice rather than left as an unverified guess, closing a gap most observability dashboards leave open.
Pros
- Tracks spend across API providers and coding tools like Cursor in one dashboard, wider than a single-gateway product such as OpenRouter that only sees traffic passing through it.
- Non-proxy, SDK-based architecture never sits in the request path and does not collect prompt or response content, a lighter footprint than proxy-based gateways.
- Free tier requires no credit card, and paid plans bill on the AI spend actually managed rather than by seat, unlike most per-seat SaaS pricing.
- Reconciles SDK cost estimates against the provider's own invoice, a verification step pure observability dashboards typically skip.
Cons
- Team is 2 people, Christopher Acker and Yuta Baba, as of its Y Combinator Winter 2026 launch, so enterprise support depth and roadmap velocity are unproven at scale.
- No public SOC 2, ISO 27001, GDPR, or HIPAA attestations were found on its site or terms as of August 2026, a gap for regulated buyers.
- Per-request attribution requires wrapping provider clients in code via the SDK; teams that want zero-integration proxy tracking get less granular tagging without that step.
Frequently Asked Questions
What does SuperPenguin actually cost?
SuperPenguin bills on managed AI spend, not seats. Free covers up to $2,000 a month, Growth is $30/month for up to $5,000 in spend and 3 team members, Pro is $200/month for up to $20,000 and 10 team members, and Enterprise is custom-priced above $20,000 with unlimited team members.
Can you use SuperPenguin without paying?
Yes, the free plan requires no credit card and includes both attribution SDKs plus the Mac menu bar app for per-pull-request coding cost. It stops syncing new data once the account's managed AI spend crosses the free plan's monthly cap, though historical dashboards stay viewable.
What are SuperPenguin's closest competitors?
The closest comparisons are LLM gateways and observability tools such as OpenRouter, Helicone, and Portkey. Choose one of those instead if a proxy-based unified API endpoint, built-in caching, or request routing matters more than SDK-based, non-proxy attribution across both API spend and coding-agent tools like Cursor.
How does SuperPenguin compare to OpenRouter in 2026?
OpenRouter routes requests through a single API and can only see the traffic that actually passes through it. SuperPenguin instead reads cost data through lightweight SDKs, covering spend across every connected provider a team uses, OpenRouter included, and checks its numbers against each provider's real invoice.
How do you set up SuperPenguin?
Create an account with no credit card required, then connect a provider by API key for an automatic, zero-code cost sync with model-level breakdowns. For per-request attribution, install the Python or TypeScript SDK, wrap the existing provider client, and add a customer or feature tag to each call.
Top Alternatives
- OpenRouter: Pick SuperPenguin if you want non-proxy cost attribution across every provider you already use; pick OpenRouter if you want a single routing endpoint that unifies model access behind one API.
- Crukx: Pick SuperPenguin if you need spend attribution by customer and PR-level coding cost; pick Crukx if you need deeper enterprise LLM observability and optimization tracing for agents in production.
- Touchmark: Pick SuperPenguin if you want visibility into spend across every AI provider you use; pick Touchmark if you want to adjust what customers are billed based on per-output quality scores.