Anthropic Target

Test Anthropic's Claude models.

Setup

akira> target anthropic https://api.anthropic.com/v1/messages -k $ANTHROPIC_API_KEY

Options

Option	Description	Default
`-k, --key`	Anthropic API key	Required
`-m, --model`	Model ID	`claude-3-sonnet-20240229`

Model Selection

akira> target anthropic https://api.anthropic.com/v1/messages \
    -k $ANTHROPIC_API_KEY \
    -m claude-3-opus-20240229

Available models:

claude-3-opus-20240229
claude-3-sonnet-20240229
claude-3-haiku-20240307
claude-3-5-sonnet-20241022

Example Session

akira> target anthropic https://api.anthropic.com/v1/messages \
    -k $ANTHROPIC_API_KEY \
    -m claude-3-opus-20240229
[+] Target configured: anthropic

akira> use dos/magic_string
[+] Using dos/magic_string
    Severity: HIGH

akira> check
[*] Checking if target is Claude...
[+] Target confirmed as Claude model

akira> run
[*] Running magic_string...

Claude-Specific Modules

Some modules are designed specifically for Claude:

dos/magic_string - Tests Claude-specific DoS vulnerability

These modules may have reduced effectiveness against other providers.

API Differences

Anthropic's API differs from OpenAI's:

Uses messages array format
Requires anthropic-version header
Different response structure

Akira handles these differences automatically when using the anthropic target type.

Keyboard shortcuts