AI Gateway & Cost Control Plane

One gateway between your apps and every AI provider. Track tokens, enforce budgets, and correlate agent costs — instead of stitching together proxies, loggers, and spreadsheets.

Get Started See How It Works GitHub Repo

Stop guessing what AI costs. Start controlling it.

01OpenAI-compatible proxy to every major AI provider — change your base URL, keep your code.→ 02Real-time cost tracking per request, model, team, and agent task.→ 03Budget enforcement with warnings, automatic throttling, and hard blocks — before the bill surprises you.→ 04Agent task correlation — orchestrators, subagents, tool calls, one total cost.→
APPLICATIONAPI Key AI AGENTMCP / Headers SAAS PLATFORMPer-customer keys CI / CDAutomated runs WARDISAI GATEWAYTrack · Budget · Route · Cache OPENAIGPT-5.4, o3 ANTHROPICOpus 4.7, Haiku GOOGLEGemini 3.1 Pro AWS / AZUREBedrock, Ollama
today's cost by model
gpt-5.44,812 req18.2M tok$842.50 claude-opus-4.73,240 req9.7M tok$485.20 haiku-4.54,150 req5.1M tok$25.60 gemini-3.11,520 req6.8M tok$381.40 sonnet-4.6487 req2.1M tok$112.60
Today14,209 requests$1,847.30

Real-time cost tracking across every provider.

Every request through Wardis is logged with token counts, cost calculations, and latency metrics. See exactly what each team, model, and agent spends — down to the individual request.

Real-time cost per request, model, team, and agent task.

Query millions of events instantly — costs, tokens, latency, all searchable and filterable.

Native support for OpenAI, Anthropic, Google, AWS, Azure, Ollama — plus any OpenAI-compatible provider.

OpenAI-compatible API — change your base URL, keep your code.

Budgets that enforce themselves.

Budget hierarchy from org to team to individual API key. Automatic warnings at 80%, throttling at 95%, and hard blocks at 100%. Sub-millisecond enforcement on every single request.

Org, team, and per-key budget limits with automatic enforcement.

Budget-aware routing automatically switches to cheaper models near limits.

Email, Slack, and webhook alerts before budgets run out.

Cost spike detection — alerts if spend jumps 200% in 5 minutes.

budget-config.yaml
org_budget: $10,000/month
team_engineering: $5,000
key_prod: $3,000
key_staging: $500
team_support: $2,000
alerts:
warn: 80% → email, slack
throttle: 95% → switch to haiku
block: 100% → reject request
task breakdown
Task: "Refactor auth module"$4.72
Orchestrator (gpt-5.4)$1.35
├─ Code analysis (claude-opus-4.7)$2.18
├─ Test generation (gpt-5.4)$0.96
└─ Documentation (haiku-4.5)$0.23
247 requests · 14 min · 4 agents

Full cost visibility across your AI agents.

A single AI agent task can spawn dozens of subagents, each making hundreds of LLM and tool calls. Wardis tracks the entire agent tree — orchestrator, subagents, tool calls — and rolls it all up into one detailed cost breakdown per task.

Full agent tree view — orchestrator, subagents, and tool calls with cost per node.

Works with Claude Code, LangChain, CrewAI, and any framework that sends headers.

Parent-child cost roll-up across unlimited nesting levels.

Built-in MCP server — Claude Code agents get automatic cost tracking with zero setup.

Built for every team

Start where it fits your work.

An open source AI gateway and cost control plane — free to self-host, with Pro and Enterprise options for teams that need more.

Developer

Self-host on any machine with a single command. Full gateway, dashboard, and agent tracking with zero vendor lock-in.

Quick Install →Documentation →

SaaS Builder

Track AI costs per customer. Map teams to customers, assign a key per customer, set per-customer budgets — Wardis handles the rest.

SaaS Use Case →Plans →

Enterprise

SSO/SAML, full audit trail, RBAC with four roles, and priority support. Built for compliance from day one.

Enterprise Features →Contact Us →