In Development

Citadel OS

Citadel captures everything. The OS surfaces everything. One interface to query every request, every cost, every decision across every AI provider your organization touches.

A query layer and operational dashboard built on top of Citadel's telemetry data. Ask a question, get an answer backed by real usage data. No digging through logs.

citadel-os
$ citadel query "who spent the most on AI last month?"
# Result:
team: engineering
spend: $8,420.33
models: claude-sonnet-4-6 (62%), gpt-4o (28%)
requests: 34,291
trend: +12% vs prior month

Citadel

Captures every AI request, cost, and provider interaction. The data layer.

Learn more →

Citadel OS

Surfaces everything Citadel captures. The query and decision layer.

You are here
Capabilities

Ask. See. Decide.

Every question about your AI operations has an answer. Citadel OS finds it.

Natural Language Queries

ask in plain english

Ask "which team is burning the most on AI?" and get an answer with numbers. No SQL. No dashboard hunting. The OS interprets your question against Citadel telemetry and returns a direct answer.

Cost Center Dashboard

spend by team, project, model, provider

Every dollar tracked to the team, project, and model that spent it. See trends over time, spot anomalies before they become surprises, and enforce budgets at the organizational level.

Usage Analytics

who is using what and how often

Request volume, model distribution, latency percentiles, error rates. If it passed through Citadel, the OS can slice it. Filter by team, time range, provider, or model.

Audit Trail Explorer

full request and response history

Search the complete history of AI interactions. Find the exact request that caused a cost spike. Trace a model response back to the user who triggered it. Every field is searchable.

Queries

Questions You Can Answer

These are not hypothetical. If Citadel is capturing the data, the OS can surface it.

How much did we spend on Claude vs GPT-4o last quarter?

Which project has the highest cost per request?

Show me all requests from the marketing team in the last 7 days.

What percentage of our AI budget is going to embedding models?

Flag any team that exceeded 80% of their monthly budget.

What is our average latency to Anthropic vs OpenAI?

Which model gives us the best cost-to-quality ratio for summarization?

citadel-os query
# Cost breakdown by provider, last 30 days
$ citadel os cost-breakdown --period 30d

provider      requests   spend
anthropic    48,201     $6,840.12
openai       22,405     $3,210.44
google       8,102      $920.88

total:       78,708     $10,971.44
avg/request: $0.14

# Flag: engineering team at 91% of monthly budget

You Will Hate This If

  • You want a pretty dashboard that nobody queries
  • You do not have Citadel capturing data yet

See what your AI operations actually look like.

Citadel OS is in early access. If you are already running Citadel or want to start with both, get in touch.

Request Early Access