generated: '2026-08-02'
method: searched
source: https://docs.flintai.dev/flintai/cli/reference/commands
description: 'SandboxAQ ships two first-party command-line tools across two products: flintai, the Flint
  AI agent-security CLI (scan + eval), and cs-api, the AQtive Guard API client that wraps trace upload
  and analysis. The binaries themselves are catalogued in packages/sandboxaq-packages.yml; this file is
  their command surface.'
clis:
- name: flintai
  product: Flint AI
  package: flintai-cli
  docs: https://docs.flintai.dev/flintai/cli/index
  reference: https://docs.flintai.dev/flintai/cli/reference/commands
  repository: https://github.com/sandbox-quantum/flintai-cli
  changelog: https://docs.flintai.dev/flintai/cli/resources/changelog
  requirements:
  - Python 3.11 or later
  - OpenGrep (required for Flint AI Scan)
  - A running agent accessible over HTTP (required for Flint AI Eval)
  install:
  - method: pip
    command: pip install flintai-cli
  config_files:
  - path: ~/.flintai/config.json
    purpose: Eval models, evaluations and model-evaluation assignments
  - path: ~/.flintai/.env
    purpose: LLM provider API keys
  global_options:
  - {flag: --version, description: Print the installed version and exit}
  commands:
  - name: flintai init
    description: 'Interactive setup — select an LLM provider (Gemini, OpenAI, Anthropic or LiteLLM),
      choose a model and enter an API key. Creates ~/.flintai/.env and ~/.flintai/config.json.'
  - name: flintai scan
    description: 'AI-powered static security analysis of agent code (whitebox testing). Findings are
      mapped to the OWASP Top 10 for Agentic Applications with CVSS severity scores.'
    arguments:
    - {name: path, required: true, description: Path to a file or folder to scan}
    flags:
    - {flag: --output/-o, default: 'scan_<timestamp>.<format>', description: Output file for results}
    - {flag: --format/-f, default: json, values: [json, sarif], description: 'Output format; SARIF for
        GitHub code scanning'}
  - name: flintai eval
    description: Runtime behavioural evaluation with adversarial prompts (blackbox testing). Scores 0-100%
      mapped to the OWASP Top 10.
    flags:
    - {flag: --config, default: ~/.flintai/config.json, description: Path to the JSON config file}
    - {flag: --log, default: 'flintai_<timestamp>.log', description: Log file path}
    subcommands:
    - {name: flintai eval models list, description: 'List configured models/agents (supports --tag)'}
    - {name: flintai eval models show, description: Show details for a model by full ID or unique prefix}
    - {name: flintai eval evaluations list, description: 'List all evaluations, builtin plus user
        (supports --tag)'}
    - {name: flintai eval evaluations show, description: Show evaluation details and connected models}
    - {name: flintai eval model-evaluations list, description: List model-evaluation assignments}
    - {name: flintai eval model-evaluations attach, description: 'Attach evaluations to models by ID or
        by tag (--model, --eval, --model-tag, --eval-tag)'}
    - {name: flintai eval model-evaluations detach, description: Remove assignments by ID or by tag}
    - name: flintai eval run
      description: Run evaluations for a model or a specific model-evaluation ID.
      flags:
      - {flag: model_evaluation_id, description: 'Optional positional; omit and select with --model or
          tag filters instead'}
      - {flag: --model, description: Run all evaluations assigned to this model ID}
      - {flag: --output/-o, default: 'eval_<timestamp>.<format>'}
      - {flag: --format/-f, default: json, values: [json, sarif]}
      - {flag: --concurrency/-c, default: 20, description: Max concurrent evaluation tasks}
      - {flag: --model-tag, description: Filter by model tag (repeatable)}
      - {flag: --eval-tag, description: Filter by evaluation tag (repeatable)}
  key_flows:
  - name: Scan quickstart
    steps: [pip install flintai-cli, flintai init, flintai scan /path/to/agent/code]
    docs: https://docs.flintai.dev/flintai/cli/guides/scan-quickstart
  - name: CI/CD integration
    description: 'Emit SARIF (--format sarif) from scan or eval and upload to GitHub code scanning;
      documented for GitHub Actions, GitLab CI and CircleCI.'
    docs: https://docs.flintai.dev/flintai/cli/guides/ci-cd-integration
  supported_agent_frameworks:
  - Google ADK
  - Google GenAI
  - Anthropic
  - OpenAI
  - OpenAI Agents SDK
  - LangGraph
  - CrewAI
  - AutoGen
  - HuggingFace Transformers
  - HuggingFace smolagents
- name: cs-api
  product: AQtive Guard
  docs: https://aqtiveguard.sandboxaq.com/docs/api/api-client/
  reference: https://aqtiveguard.sandboxaq.com/docs/api/api-client/manual/
  installation: https://aqtiveguard.sandboxaq.com/docs/api/api-client/installation/
  repository: https://github.com/cryptosense/api-client
  description: CLI program that interfaces with AQtive Guard through its GraphQL API — uploads trace
    files and runs analysis without hand-writing GraphQL.
  authentication:
    flag: -k, --api-key <string>
    env: CRYPTOSENSE_API_KEY
  commands:
  - {name: cs-api --analyze, description: Upload and analyze traces}
  - {name: list-profiles, description: Retrieve available profile IDs}
  - {name: upload-trace, description: Upload trace files}
  flags:
  - {flag: -u, --api-base-url <url>, default: 'https://aqtiveguard.sandboxaq.com', description: 'API base
      URL; change it for on-premise deployments'}
  - {flag: --ca-file <path>, description: PEM certificate bundle for certificate validation}
  - {flag: --no-check-certificate, description: Disable certificate checking}
  proxy_support: http_proxy / https_proxy environment variables