CostHawk
Repos, and Projects
Last verified:
What is CostHawk?
CostHawk is an AI API cost monitoring and adoption tracking platform built for engineering teams. It provides real-time visibility into AI API spend across every provider, model, and project, consolidating data from OpenAI, Anthropic, Google, xAI, Mistral, DeepSeek, AWS Bedrock, Azure OpenAI, and 5+ other providers into one unified dashboard. The tool was created by Chase Dillingham, a full-stack engineer who built it after his team at TrainMyAgent was blindsided by unexpected API bills that were 2-3 times higher than expected.
Key features include real-time cost tracking with budget alerts, adoption monitoring showing who on the team is using AI and how often, wrapped proxy keys for routing API calls through CostHawk for attribution and cost limits, MCP server for local-first telemetry from Claude Code, OpenAI Codex CLI, Cursor, and Gemini CLI, admin API connections for automatic provider data sync, cost optimization tools to find cheaper models, savings engine comparing subscription vs pay-per-use plans, finance-ready data export in CSV/JSON formats, REST API for programmatic access, and private team benchmarks scoped to your organization with role-based access control. The platform tracks 196 models with live pricing data and supports 13 AI provider families.
CostHawk is designed for engineering leaders, CTOs, CPOs, and CFOs who need to prove AI adoption, control AI spend, and connect AI investment to roadmap throughput. Engineering leads can spot champions and stalled teams, CTOs get one operating view across teams/tools/providers, CPOs connect AI to product shipping, and CFOs track cost before it sprawls. The tool is also useful for product and finance leaders who need visibility into AI usage without building internal reporting systems from scratch.
CostHawk pricing
Pricing model: Freemium
CostHawk offers a 30-day free trial on both Individual and Team plans with credit card required at signup but no charge until the trial ends. You can cancel anytime from the billing portal. Paid plans begin at $19/month. Individual plan includes: personal AI adoption dashboard, Claude Code/Codex/Cursor sync, desktop MCP connector, and email support. Team plan (most popular) includes everything in Individual plus: team adoption roster, project and team membership, and owner/admin org reporting. Enterprise plan includes everything in Team plus: custom integrations, SSO/SCIM, and dedicated support with SLA. Enterprise is for organizations with 101+ people or custom integrations/support and requires contacting sales at [email protected]. Pricing is per seat where a seat is each unique team member with workspace access. They do not bill per AI session, per token, or per integration. Plan changes prorate and can be made from the billing portal at any time.
CostHawk pros
- Real-time visibility into AI API spend across all providers
- One dashboard for every AI provider instead of scattered billing portals
- Local-first MCP telemetry without needing admin credentials to start
- Supports 13 AI provider families including OpenAI, Anthropic, Google, xAI, Mistral, DeepSeek
- Tracks 196 models with live pricing data indexed
- 30-day free trial with card required but no charge until trial ends
- Budget alerts notify before costs exceed limits
- Wrapped proxy keys isolate real provider keys from codebase
- AES-256 encryption for admin API keys at rest
- Never stores prompt text, responses, source code, or conversation history
- 5-minute setup with no infrastructure changes required
- Private team benchmarks scoped to organization with role-based access control
- Cost optimization recommendations to find cheaper models without quality loss
- Finance-ready CSV/JSON export for invoicing and reporting
- REST API for programmatic access to adoption and cost data
CostHawk cons
- Credit card required at signup for trial
- Paid plans start at $19/month which may be pricey for solo developers
- Relative new tool without long track record
- Some features like Provider Comparisons still coming soon
- MCP server support for OpenCode listed as coming soon
- Enterprise tier requires contacting sales no transparent pricing
- Bill shows provider cost endpoints are daily by design not real-time for costs
- Limited to 13 provider families some niche providers not supported
- Team benchmarks only visible to authorized members may limit transparency
- Auto-sync disabled by default requires manual configuration enable
Frequently asked questions about CostHawk
What is CostHawk?
CostHawk is an AI adoption monitoring platform for engineering teams. It tracks who on your team is using AI tools, how much they cost, and how adoption is trending across OpenAI, Anthropic, Google, and other providers. It provides adoption, cost, and impact across AI tools in one secure view.
How does CostHawk track AI adoption across my team?
CostHawk supports three integration methods: MCP telemetry for local developer activity, admin API sync for organization-wide provider cost data, or wrapped proxy keys that sit between your code and the provider. Install the MCP with one command to authenticate your machine and connect Claude Code, Codex, Cursor, and Gemini CLI.
How are team benchmarks kept private?
Team Benchmarks are scoped to your organization and protected by role-based access control. Only owners, admins, and authorized members can view rankings. Workspace member names stay inside the private internal view. The public leaderboard uses aliases only and never uses real names, account names, or emails.
What providers does CostHawk support?
CostHawk supports OpenAI, Anthropic, Google, xAI, Mistral, Cohere, DeepSeek, Groq, Meta, Perplexity, Together, AWS Bedrock, and Azure OpenAI. They add new providers regularly and currently support 13 AI provider families with 196 models tracked.
What data does CostHawk see — and not see?
CostHawk stores only telemetry metadata: token counts, model names, timestamps, and hashed project IDs. They never see, store, or process your prompts, responses, or code. They do not collect prompt or completion content, source code, file contents, plaintext API keys, or conversation history.
How do I get started?
Sign up, invite your team, and connect via MCP telemetry or admin API keys. The entire setup takes under 5 minutes. Run npm exec --yes costhawk@latest -- --login to open your browser, approve login, and write MCP config automatically. Card is required at signup but you won't be charged until the 30-day trial ends.
Why is a credit card required for the trial?
It keeps abuse low and means there's no extra step on day 31. You can cancel any time during the trial and you won't be charged. On day 30 they charge the card for your tier if you haven't cancelled.
What happens when the trial ends?
On day 30 they charge the card for your tier (Individual or Team). If you cancel before then, no charge. If the card fails, they retry the payment before access is paused. You can cancel anytime from the billing portal.
Can I switch plans later?
Yes. You can change tiers from the billing portal at any time. Changes prorate so you only pay for the portion of the billing period at each tier level.
What counts as a seat?
Each unique team member with access to your CostHawk workspace counts as a seat. They don't bill per AI session, per token, or per integration — just seats. For Enterprise with 101+ people, contact [email protected] for custom pricing.