AIHelms

AIHelms is an enterprise AI resource governance platform from 北京明睿博通科技有限公司, published as an open-source project under GPL-3.0. It lets administrators connect model providers, issue AI identities (API keys) to employees, and track token-level cost and usage. The latest release listed is 0.1.25 (2026-09-26).

What It Is

AIHelms is a management layer for enterprise AI use, built around three audiences: decision makers who see usage, cost and reports; administrators who connect models, assign identities and set budgets; and employees who obtain an AI identity and use it from any client. It is deployed with Docker Compose and exposes a user-facing AI Hub, an admin console and API docs.

Cost Control and Identity

The platform separates external cost (what suppliers charge) from internal settlement prices charged to departments. Budgets can be set per person, department, project or model, with soft limits (warnings) and hard limits (suspension on overrun). Each call is attributed to a person, department, project and model down to token level. Each employee gets one AI identity (API key) as a single entry point, with rate limits and model whitelists.

Model Management and AI Market

Administrators register multiple providers, manage credentials, load-balance several deployments of the same model, and test connectivity. Both OpenAI and Anthropic API formats are supported. Skills and MCP servers can be registered, approved and distributed internally, and employees can request access through the AI Market. Security features include admin audit logs, and AI Policies scanning of Skills for risks such as malicious instructions or data exfiltration.

Efficiency Analytics

Dashboards cover department coverage, user activation, model usage distribution, daily calls, per-person token use, cost rankings and budget progress. Weekly and monthly efficiency reports can be drilled down by department, project or model.

Setup Path

Deployment requires Docker 20.10+ with Compose v2 and at least 4 cores and 8 GB RAM. After cloning and configuring .env, run docker compose up -d. Then add provider credentials, create and publish a model, and copy the key and endpoint into a client. The stack uses FastAPI, Vue 3, PostgreSQL and Redis, with LiteLLM community edition as the base layer. An enterprise edition adds official support and identity integrations such as LDAP, DingTalk and SSO.