# AI Command Center

> A dependency-free LLM gateway and self-hosted usage & cost dashboard. Point any
> project at it (any language, one command) and it captures tokens, cost, latency,
> and errors per project. Provider API keys pass through untouched; prompt and
> response bodies are never stored - metadata only.

Install and run:
- `npx ai-command-center` starts the gateway + dashboard at http://localhost:4321
- `npx ai-command-center demo` seeds 14 days of realistic sample data
- Node.js >= 18.17 is the only prerequisite; the gateway has zero runtime dependencies

Integrate (change a base URL, keep your API key):
- OpenAI-compatible: `http://localhost:4321/p/<project>/openai/v1`
- Anthropic: `http://localhost:4321/p/<project>/anthropic`
- Gemini: `http://localhost:4321/p/<project>/gemini`
- Also: openrouter, mistral, deepseek, xai, groq, together, ollama, and custom OpenAI-compatible endpoints
- Or set the SDK's standard env var (OPENAI_BASE_URL / ANTHROPIC_BASE_URL / GOOGLE_GEMINI_BASE_URL) - zero code change
- Group calls by project via the `/p/<project>/` path segment or an `x-aicc-project` header
- With auth enabled the path becomes `/k/<gateway-key>/…` (the key authenticates and sets the project)
- Group a request/agent run into a session with an `x-aicc-trace` header; track prompts with `x-aicc-prompt` / `x-aicc-prompt-version`
- Report usage the proxy can't see with `POST /api/track` (JSON: project, provider, model, tokensIn, tokensOut, trace?, prompt?, promptVersion?)

Dashboard views: Overview (cost/usage/latency), Traces (sessions), Prompts (version metrics), Models (comparison), Alerts (budgets + anomalies).

What it is: self-hosted, language-agnostic dashboard for cost, usage, traces, prompt versions, budgets and anomaly alerts; a transparent single-upstream proxy.
What it is NOT (by design): a distributed span-tree tracer, an LLM-as-judge eval framework, a prompt playground, or a routing/failover gateway.

## Docs
- Overview: https://aicommandcenter.vercel.app/docs
- Install & run: https://aicommandcenter.vercel.app/docs/install
- Integrate a project: https://aicommandcenter.vercel.app/docs/integrate
- Traces, prompts & budgets: https://aicommandcenter.vercel.app/docs/features
- Configuration: https://aicommandcenter.vercel.app/docs/config
- Auth & teams: https://aicommandcenter.vercel.app/docs/auth
- Providers & pricing: https://aicommandcenter.vercel.app/docs/providers
- Self-hosting: https://aicommandcenter.vercel.app/docs/self-hosting
- Security: https://aicommandcenter.vercel.app/docs/security
- HTTP API: https://aicommandcenter.vercel.app/docs/api
- Comparison with other tools: https://aicommandcenter.vercel.app/docs/comparison
- FAQ: https://aicommandcenter.vercel.app/docs/faq

## Source
- GitHub: https://github.com/adityasarade/ai-command-center
- npm: https://www.npmjs.com/package/ai-command-center
- License: MIT