Enterprise AI gateway with an accounting department

What it does

AIHelms is an enterprise management layer built atop LiteLLM that turns scattered AI usage into accountable line items. It provides a unified gateway for OpenAI, Azure, Claude, DeepSeek and others, while a Vue-based admin dashboard handles model registration, API key lifecycle, and budget quotas. Employees get a self-service portal to copy keys and track their own burn rate.

The interesting bit

The standout feature is genuinely granular cost governance: external vendor pricing and internal transfer pricing run on separate tracks, so you can mark up model costs when billing departments or pass them through at cost. Budgets can be hard caps that cut off access or soft warnings that nag politely, with every call traceable to a specific person, project, and model down to the token.

Key highlights

• Dual-track pricing lets finance reconcile vendor bills against internal cost centers without spreadsheet archaeology.
• Token-level cost attribution traces spend to individual users, departments, and projects.
• Built-in marketplace for MCP servers and internal Skills, with approval workflows.
• Admin dashboard covers model connectivity testing, load balancing across multiple deployments, and audit logging.
• Built for Docker Compose with FastAPI backend, Vue 3 frontend, PostgreSQL, and Redis.

Caveats

• Explicitly built on LiteLLM community edition, so the core gateway is not homegrown; this is primarily a management and billing orchestration layer.
• GPL-3.0 licensed, which the authors acknowledge may scare off corporate legal teams; commercial licenses are offered by email.
• Several roadmap items (AI employees, A2A agent collaboration, context caching) are marked as work-in-progress or future plans rather than shipped features.

Verdict

Worth a look if your company runs multiple LLM contracts and needs to prove ROI to finance. Skip it if you want a lightweight proxy; this is a full-stack management platform that expects to own the stack.