Core concepts

BotGauge works with your agent in two different ways, and each has its own vocabulary. This page defines the terms so the rest of the docs read easily.

Two ways BotGauge works with your agent

	Red teaming	Tracing
What happens	BotGauge sends your agent attack messages and reads its replies	Your app sends spans to BotGauge as it runs
How it connects	Your agent’s endpoint URL, plus any secrets it needs	The Python SDK, or any OTLP exporter, with a project API key
What you learn	Which rules the agent broke, held, or couldn’t be tested on	The prompts, tool calls, and execution path behind a result
Start here	Red teaming quickstart	Tracing quickstart

The word “agent” has two meanings. In red teaming it is the AI system under test. In tracing it is the named agent you mark with tracing.agent(), which labels the spans that agent produced.

How your team is organized

An organization has many workspaces, and each workspace has many projects. Projects hold agents and logs, and they have their own secrets and API keys. Roles exist at the organization and workspace levels. See Organizations, workspaces, and projects.

Red teaming terms

Term	Meaning
Agent	The AI system you are testing
Connection test	The quick check that BotGauge can reach your agent and read its reply
Recon	The discovery chat where BotGauge learns what your agent does
Rule	A statement of what your agent must or must not do
Plan	The attacks that will run for the selected rules
Approval	Your sign-off on a specific plan, with limits and an expiry
Run	One execution of an approved plan
Finding	A recorded problem, from a confirmed breach or a signal
Held	The agent resisted the attacks. The only passing result.
Signal	Suspicious but not proven
Not evaluable	Couldn’t be tested here, for example because of missing access
Black box, grey box	Testing from outside only, and testing with inside visibility
Verify fix	Replaying a finding’s attack to confirm the fix

The full list is in the red teaming glossary.

Tracing terms

Term	Meaning
Span	One timed operation, such as an LLM call or a tool call
Trace	The set of spans that make up one request
Session	The long-lived conversation or run, set with `session_id`
Turn	One unit of work inside a session, opened with `tracing.turn()`
Agent (in tracing)	The named agent doing the work, opened with `tracing.agent()`
Instrumentor	The code that records one framework’s LLM and tool calls as spans
Extra	An optional install, such as `botgauge[langchain]`, that adds an instrumentor
Capture	The setting that selects which content is exported: `metadata`, `text`, or `full`
Redaction	The setting that decides whether detected PII and credentials are masked
OTLP	The OpenTelemetry protocol BotGauge accepts at `https://api.botgauge.com/v1/traces`
Logs	The console page, under Observability, where your sessions, traces, and spans appear
Judge	An evaluator that scores your logs, for example Answer Relevance. See Built-in judges.

Explore red teaming and tracing

Now that you know the core concepts, sign in to BotGauge and set up your first project.

Start here

Two ways BotGauge works with your agent

How your team is organized

Red teaming terms

Tracing terms

Where to go next

Set up your account

Red-team an agent

​Two ways BotGauge works with your agent

​How your team is organized

​Red teaming terms

​Tracing terms

​Where to go next

Set up your account

Red-team an agent

Two ways BotGauge works with your agent

How your team is organized

Red teaming terms

Tracing terms

Where to go next