OpenA2A: CLI · HackMyAgent · Secretless · AIM · Browser Guard · DVAA

opena2a

Open-source security platform for AI agents. Installed as opena2a-cli on npm.

npx opena2a-cli review

  OpenA2A Security Review  v0.8.21

  Findings
  -----------------------------------------------
  Credential scan        3 hardcoded keys
  Shadow AI              2 agents, 4 MCP servers
  Config integrity       unsigned
  Governance             no SOUL.md
  -----------------------------------------------
  Security Score   30 / 100  -> 85 by running opena2a protect

  Run: opena2a protect    (fix all findings)

Install globally if you prefer:

npm install -g opena2a-cli
brew tap opena2a-org/tap && brew install opena2a

Built-in Help

You do not need this README. The CLI has built-in discovery:

opena2a ?                           # Contextual recommendations for your project
opena2a ~shadow ai                  # Semantic search across all commands
opena2a "find leaked credentials"   # Natural language command matching
opena2a                             # Interactive guided wizard (no args)

Commands

| Command | What it does | |---------|-------------| | opena2a setup | One-command onboarding — auth, identity, MCP discovery, trust score | | opena2a review | Full security dashboard — HTML report, 6-phase assessment | | opena2a detect | Find shadow AI agents, MCP servers, AI configs. Governance score. | | opena2a detect --report | Executive HTML report | | opena2a detect --export-csv | Asset inventory for CMDB/ServiceNow | | opena2a init | Read-only security assessment with trust score | | opena2a protect | Fix everything — credentials, .gitignore, config signing | | opena2a watch | Live tail of agent activity events | | opena2a identity create | Cryptographic identity for your project | | opena2a identity mcp attach | Auto-discover and attach MCP servers | | opena2a identity integrate | Wire security tools to identity (audit + trust) | | opena2a harden-soul | Generate SOUL.md governance rules | | opena2a scan | 204 security checks via HackMyAgent | | opena2a mcp audit | Audit MCP server configurations with trust scores | | opena2a guard sign | Sign config files for tamper detection | | opena2a shield init | Full security setup — all of the above, one command |

Ecosystem

Each command routes to a specialized tool, installed on first use:

| Command | Tool | Description | |---------|------|-------------| | detect | Shadow AI | Discover AI agents, MCP servers, AI configs | | identity | AIM | Cryptographic identity, audit logs, trust scoring | | scan | HackMyAgent | 204 security checks, attack simulation, auto-fix | | secrets | Secretless AI | Credential management for AI coding tools | | mcp | MCP Security | Audit, sign, and verify MCP server configurations | | benchmark | OASB | 222 attack scenarios, compliance scoring | | train | DVAA | Vulnerable AI agent for security training |

Use Cases

• Developer using AI coding tools — 5 minutes
• Security team assessing AI risk — 10 minutes
• MCP server author — 15 minutes
• CI/CD pipeline integration

Docs

Full command reference, Shield subcommands, scope drift detection, behavioral governance, credential patterns, and CI/CD examples: opena2a.org/docs

Requirements

• Node.js >= 18
• Optional: Docker (for opena2a train)

Telemetry

opena2a sends anonymous tier-1 usage data to the OpenA2A Registry: tool name, version, command name (check, scan, protect, etc.), success, duration, platform, Node major version, and a stable per-machine install_id. No content is collected — no scanned packages, no findings, no file paths, no env-var values, no IPs.

This is separate from the community-contribution data the Registry already collects via --publish flags; that flow continues to populate the public community page (scan submissions, findings, contributors). Telemetry just answers "is anyone running these commands at all?"

• Policy: opena2a.org/telemetry.
• Status: opena2a telemetry status.
• Disable per-invocation: OPENA2A_TELEMETRY=off opena2a <anything>.
• Disable persistently: opena2a telemetry off.
• Audit every payload: OPENA2A_TELEMETRY_DEBUG=print opena2a <anything> echoes each event to stderr in JSON.

Fire-and-forget with a 2-second timeout — telemetry never blocks the CLI.

License

Apache-2.0

Website · Docs · Discord · GitHub