Dashboard walkthrough: every page in the Cognocient AI app, and which plan it needs
A tour of all 38 pages in the Cognocient AI app, in sidebar order (Observe, Monitor, Control, Intelligence, Executive): what each shows, the plan it needs, and where to start.
The app sidebar has five sections. This page walks through every page in each, in sidebar order, with the plan it needs and a link to its full guide.
Plan labels
Every plan means Free and up. Base, Growth and Business mean that plan and above. The sidebar shows the same label next to any page your plan can't open, and the 10-day trial opens everything. Full comparison: Pricing.
| Section | Pages |
|---|---|
| Observe | Overview, Workstreams, Cost per Outcome, Customer Margin, AI Coding Tools, Environments, Saved Views |
| Monitor | Live Calls, Anomalies, Compromise Risk, Waste, Token Maxing, Loop Breaker, Run Budgets, Agent Workflows |
| Control | Budgets, CI Cost Gate, Shadow Spend, Routing Rules, Auto Router, Load Balancer, Integrations, MCP Gateway, Security & SSO, Audit Log, API Keys, Safety & Limits, Chargeback |
| Intelligence | Recommendations, Feature Intelligence, Experiments, AI Advisor, Insights, Forecast, Cost Simulator, Reports |
| Executive | Executive View, Outcome Tracking |
Observe
Where your AI spend goes.
Overview — /dashboard · every plan

The page you land on. It has three tabs: Overview, Sessions and Users (the last two need Growth).
- Get set up — a checklist that ticks itself off from your account's real state: connect a provider, send a first call, tag calls by feature, set a budget, separate production from test, and more. It disappears when every item is done, or when you hide it.
- KPI cards — Total Calls Today, Cost Today, Error Rate and Retry Waste, each against yesterday. Retry waste is spend on calls that failed and were retried; it is usually the fastest saving.
- System Health — average Latency, Budget Usage against your limits, the Circuit Breaker (trips when a key's token rate jumps far above its baseline) and Velocity.
- Charts — calls and cost over the last 7 days, cost by model, and spend by feature. Spend by department needs Base.
- FinOps Maturity score (Growth) and, once features are classified, a Feature Intelligence summary of investment vs waste (Growth).
- Pinned views — charts your team pinned from Saved Views (Growth).
- A banner when anomalies are open, and on the Free plan a banner showing how close you are to the $50/month tracked-spend cap.
Guide: Attribution headers · Sessions · Per-user cost
Workstreams — /dashboard/workstreams · Growth
Cost per conversation, ticket, pull request or agent task, grouped by X-Cost-Session. Use it to answer "what did this JIRA story cost in AI calls?" Guide: Sessions & Workstreams
Cost per Outcome — /dashboard/outcomes · Growth
Spend divided by business results you report with X-Cost-Outcome: cost per resolved ticket, per document processed, per conversion. Guide: Outcomes & ROI
Customer Margin — /dashboard/margin · Growth
The AI cost of each of your customers (from X-Cost-Customer) next to what they pay you, from Stripe or figures you enter, with alerts when a customer's margin drops. Guide: Customer Margin
AI Coding Tools — /dashboard/coding-tools · Growth
Cursor, GitHub Copilot and Claude Code spend per developer in one view, next to the rest of your AI bill. Guide: AI Coding Tools
Environments — /dashboard/environments · every plan
Production, staging, CI and eval spend side by side, from X-Cost-Environment, so test traffic doesn't hide inside production numbers. Guide: Environments & Variants
Saved Views — /dashboard/views · Growth
Charts you build from any metric, split and filtered, shared with the team and pinned to the Overview. Guide: Saved Views
Monitor
What is happening right now.
Live Calls — /calls · every plan

The API Call Log: every call through the gateway with timestamp, model, feature, department, user, workload, tokens, cost, latency and status. Filter by date range (24 hours to 90 days), feature or department, or search by model. Export CSV needs Base.
This is the first place to look after setup: if a call isn't here, it didn't go through the gateway. Untagged calls appear under untagged. Guide: Live Call Monitor
Anomalies — /anomalies · every plan
Spend spikes, call-frequency spikes and rejection-rate spikes measured against your 14-day baseline. Each shows the normal baseline, the current value and a probable cause (agent loop, retry storm, model changed, prompt length increased, new proxy key, traffic spike). Dismiss once handled. Guide: Anomaly Detection
Compromise Risk — /dashboard/compromise-risk · Base
Flags a proxy key suddenly used from a source IP or model family it has never used, a leaked-key signal that can stay invisible to cost-based anomaly detection. Guide: Compromise Risk
Waste — /waste · Base

Total Recoverable Waste for the last 7, 14, 30 or 90 days, split into Retry Waste, Potential Model Savings (an expensive model where a cheaper one would do) and Context Window Waste (features with oversized or highly variable prompts), with the top features for each. The Context Tax tab needs Growth. Guide: Waste Detection · Context Tax
Token Maxing — /dashboard/token-maxing · Growth
Frontier-model calls that return short outputs, the tasks a smaller model handles for a fraction of the price, by feature. Guide: Token Maxing Detector
Loop Breaker — /dashboard/failure-loops · Growth
Agent runs stopped or flagged by the Failure Loop Breaker after repeated identical tool calls or consecutive tool errors, with the session, tool, reason and action taken. Set its mode here: Off, Alert (the default) or Kill, which stops the run. Detection needs X-Cost-Session on the calls. Guide: Debug runaway agent loops
Run Budgets — /dashboard/run-budgets · every plan
Every agent run in the last 24 hours (X-Cost-Run-ID) with its calls, subagents, spend against its ceiling (X-Cost-Run-Budget) and status, plus past runs for 30 days. Guide: Track cost per agent run
Agent Workflows — /agent-workflows · Business
The cost tree of a multi-agent run: which MCP server, which tool call and which handoff spent what. Built from the MCP/A2A headers. Guide: MCP / A2A Attribution
Control
Stopping and steering spend.
Budgets — /budgets · every plan

Each budget card shows spend against its limit, color-coded as it fills, and its enforcement mode:
- Alert — notify by email (every plan) or Slack (Base) and let calls through.
- Degrade — switch calls to a cheaper model you choose once the limit is hit, so the feature keeps working.
- Block — refuse calls with a
429before they reach the provider.
When creating a budget you choose:
| Setting | Options | Plan |
|---|---|---|
| Scope | Global, feature, department, user, environment | every plan |
| Proxy key, and parent budgets that cap their children | Growth | |
| One of your customers, a spend group, a group of models | Growth | |
| Period | Monthly or daily | every plan |
| Rollover | Unused budget carries into the next period, up to one period's limit | Growth |
| Fallback allowance | An extra dollar amount spent on a cheaper model after a Block budget runs out | every plan |
| Per-run limit | A ceiling for each agent run within the scope | every plan |
| Expiry | The budget ends on a date (useful for experiments) | every plan |
| If the budget store is unreachable | Fail open (recommended) or fail closed | every plan |
The Free plan allows one budget. Guide: Budget Enforcement · Set a spending limit · Spend groups
CI Cost Gate — /dashboard/cost-gate · Base
Recent CI runs compared with their baseline cost, and which ones failed the build on a cost regression. Guide: CI Cost Gate
Shadow Spend — /shadow-spend · Growth
What each provider actually billed next to what the gateway saw, per period, with alerts on the gap: spend that never went through Cognocient AI. Connect a live billing source or enter figures by hand. Guide: Shadow Spend Reconciliation
Routing Rules — /routing-rules · Growth
Rules that rewrite the model for matching calls (by feature, department, model or token count), with no code change. Applying a recommendation creates one. Guide: Routing Rules
Auto Router — /auto-router · Growth
One model name that sends each request to a simple, medium, complex or reasoning tier by its complexity, with the savings it measured against always using the top tier. Guide: Auto Router
Load Balancer — /load-balancer · Growth
Model groups that spread one model name across several models or providers by weight, rotation or lowest cost, and the API key pool for one provider. Guide: Load Balancing · Multi-key balancing
Integrations — /integrations · Base
Connections out of Cognocient AI, each with its own plan:
| Integration | Plan |
|---|---|
| MCP server — ask about your spend from Claude Code or Cursor | Base |
| Langfuse & OpenTelemetry export, Prometheus metrics | Growth |
| FinOps connectors (Vantage, CloudZero, S3), secret managers | Business |
MCP Gateway — /mcp-gateway · Business
Your MCP servers behind the gateway: tool allow and deny lists, argument scanning, per-call pricing and attribution for every agent tool call. Guide: MCP Gateway
Security & SSO — /settings/security · Business
SAML single sign-on with verified domains, SCIM provisioning and JWT authentication. Visible to roles that can manage SSO. Guide: SSO · SCIM · JWT auth
Audit Log — /audit-log · Business
Who changed what and when, including denied attempts; searchable and exportable. Visible to roles that can view it. Guide: Audit Log
API Keys — /settings?tab=keys · every plan
Your proxy keys (sk-cog-...): create, name, revoke, and set a manual velocity limit per key. Free accounts have one key. Provider keys live under Settings → AI Providers. Guide: API Keys & Proxy Keys
Safety & Limits — /settings?tab=safety · every plan
The Emergency Freeze (stop every proxied call at once, by hand or automatically on high-risk signals) and each key's velocity protection, plus content guardrails for PII, secrets and prompt injection (Base) and provider failover (Growth). Guide: Emergency Freeze · Guardrails
Chargeback — /chargeback · Base
Spend by department mapped to your GL accounts, exported as CSV for your ERP. Guide: Chargeback & GL Mapping
Intelligence
What to change.
Recommendations — /dashboard/recommendations · Growth

The estimated monthly savings across all open recommendations, then each one with its priority, the evidence behind it, how the saving was estimated, and an action: Apply Recommendation for changes Cognocient AI can make for you (applying a model-routing recommendation creates a routing rule), or Mark as done for changes in your own code, such as moving a feature to the Batch API. Filter by Model routing, Budgets, Prompt caching or Batch API. Guide: Cut your AI bill with one click · Prompt cache & batch
Feature Intelligence — /feature-intelligence · Growth
Per-feature cost trends, unit economics and an Investment vs. Waste classification: which AI features earn their cost. Guide: Feature Intelligence · Investment vs. Waste
Experiments — /dashboard/experiments · Growth
Prompt versions or models compared side by side on real cost, tokens, latency and errors, from X-Cost-Variant. Guide: Environments & Variants
AI Advisor — /dashboard/advisor · Business
Ask questions about your spend in plain English and get answers grounded in your own call data. Guide: AI Advisor
Insights — /dashboard/insights · Growth
Findings generated each night, filterable as Warning, Opportunity, Pattern or High Urgency, each with the finding, its impact and a suggested action. Guide: AI-Generated Insights
Forecast — /forecast · Growth
30, 60 and 90-day spend projections by feature and department. What-if scenarios, which show which budgets would breach under growth, need Business. Guide: Cost Forecast
Cost Simulator — /dashboard/simulator · Business
Model how an agentic workflow will scale your spend before you roll it out, and get a budget recommendation. Guide: Agentic Cost Simulator
Reports — /reports · Base
One-click PDF reports with a written narrative (Board/CFO, Technical Finance or Engineering), budget vs actuals, and history. Scheduled delivery and FOCUS export need Growth. Guide: Board Reports · FOCUS Export
Executive
The finance view.
Executive View — /executive · Growth

Built for the CFO and the board meeting:
- Total AI Spend, vs Budget and vs Last Month.
- Spend Protected (calls blocked or degraded by budgets this month), Waste Recovered, Cache Savings (with hit rate) and the resulting Cognocient AI ROI.
- AI Efficiency Score (Business): cost per business outcome, month over month.
- AI Cost per Outcome and open Cost Anomalies.
- Generate Report for a board-ready PDF.
Guide: Executive View
Outcome Tracking — /outcomes · Growth
Set up the business outcomes you report (resolved tickets, signed contracts, conversions) and see their volume and value over time. Guide: Track cost per business outcome
Where to start
| You are | Start here | Then |
|---|---|---|
| The engineer who set it up | Live Calls — are calls and tags arriving? | Overview, then a first budget |
| Running agents | Run Budgets and Loop Breaker | Safety & Limits — know where the Emergency Freeze is |
| Looking for savings | Waste | Recommendations |
| In finance | Executive View | Chargeback and Reports |
Related articles
Platform Overview
How Cognocient AI works — proxy, attribution, enforcement, and reporting.
Use Cases
Setup guides for Customer Support, Engineering, FinOps, and CFO.
Python async wrapper
A drop-in wrapper for the OpenAI/Anthropic Python SDKs — live attribution without a proxy, honest tradeoffs included.