Core Concepts
This page explains the objects ANTS Platform works with and how they relate. Read it once before you instrument your first application.
Organizations, projects and access
| Concept | What it is |
|---|---|
| Organization | Your company or team account. Billing, SSO, data retention, custom roles and organization-wide settings live here. |
| Project | A container for data inside an organization. Every trace, API key, dashboard and setting belongs to one project. Use separate projects to separate teams or applications. |
| API keys | A public key (pk-ap-...) and secret key (sk-ap-...) pair, created per project. SDKs and OpenTelemetry exporters use them to send data. See Authentication. |
| Roles | Each member has an organization role. On plans with project roles, a project role can replace it in one project. The built-in roles are Owner, Department Admin, Security & Compliance, Cost Analyst, Manager Viewer, End User, Auditor and None. Organizations can also define custom roles. See Members & Roles. |
Your role decides which sidebar sections and pages you see. If a page in these docs is missing for you, ask an Owner to check your role.
Tracing data
| Concept | What it is |
|---|---|
| Trace | One end-to-end run of your application or agent, such as one chat turn or one task. |
| Observation | One step inside a trace. There are three types: span (a unit of work such as a tool call or retrieval), generation (an LLM call with model, tokens and cost) and event (a point in time). Observations nest to form a tree. |
| Session | A group of traces that share a session ID, such as all turns of a conversation. |
| User | The end user ID attached to a trace. Used for per-user cost and activity. |
| Score | A quality value on a trace, observation or session. Scores come from the API, from evaluators or from human annotation. Values are numeric, categorical or boolean. |
| Comment | A note on a trace, observation or session, visible to project members. |
| Environment | A label such as production or staging on every trace. Filter any view by environment. |
Read more in Observability Model and Tracing Fundamentals.
Agents
An agent is a named AI application that ANTS Platform tracks over time. The
SDKs register agents from the agent_name you set. Coding agents such as
Claude Code are registered automatically. Agents connect traces to cost,
risk, compliance and guardrail views. See
Understanding AI Agents.
How data gets in
- SDKs for Python, JavaScript and TypeScript and Java.
- OpenTelemetry (OTLP) from any OTel SDK, collector or supported tool.
- Framework integrations such as LangChain and CrewAI.
- Trace Connectors that import traces from AWS Bedrock and Google Vertex AI.
- Discover AI desktop agents and browser extensions that report AI tool usage from employee devices. See Shadow AI.
Credits
On ANTS-hosted plans, usage is measured in credits. A trace, a generation and a score each cost one credit. See Credit System.
Product modules
The sidebar groups features into modules. Each one builds on the same tracing data.
| Module | What it covers | Docs |
|---|---|---|
| Discover AI | AI tool and agent usage across the organization, plus Tracing and Sessions | Shadow AI, Observability |
| Gauge AI | Cost and latency dashboards, per-user usage, AI Optimizer and AI Inventory | FinOps |
| Budget AI | An LLM gateway with keys, budgets, limits and model routing | Budget AI |
| Governance | Risk Dashboard, AI Guardrails, AI Violations and compliance workflows | AI Governance |
Some modules need an Enterprise entitlement, a feature flag or deployment configuration. Each module's docs say what is required.