THE CLI YOUR AGENTS USE
Eight domains. One interface. The same commands work for humans and AI agents. Local or remote. Every action audited.
Agents and Humans. Same Interface.
Claude Code, GPT-4 agents, and your custom AI all use this same CLI to manage production systems. Every command returns structured JSON. Every error includes actionable codes. Every operation is auditable.
When a superagent runs systemprompt infra services restart web, it gets the same result you do. Same permissions. Same audit trail. Same interface on your laptop or a remote node.
Self-documenting commands mean agents discover capabilities at runtime. systemprompt --help --json returns complete API schemas. No hard-coded integrations required.
- Structured JSON Output — Every command returns parseable JSON with status, data, and metadata fields. Agents process results without string parsing.
- Actionable Error Codes — Errors include machine-readable codes, suggested fixes, and documentation links. Agents can programmatically handle failures.
- Runtime Discovery — --help --json returns full API schemas including parameter types, constraints, and examples. Zero hard-coded integrations.
Local to Remote in One Command
Your local CLI is a secure gateway to any SystemPrompt node. Switch profiles and every command routes transparently to the target environment. No SSH. No VPN. No credential juggling.
systemprompt cloud profile use production switches your entire session to remote execution. Inspect logs, restart services, rebuild agents—all from your local terminal with full audit visibility.
Authentication flows automatically. OAuth tokens refresh silently. The CLI handles credential rotation so you never touch secrets directly. Secure-by-default for AI infrastructure.
- Profile Switching — systemprompt cloud profile use production routes all commands to remote. No VPN, no SSH tunnels, no bastion hosts.
- Silent Token Refresh — OAuth tokens refresh automatically in the background. Sessions stay active without manual re-authentication.
- Secure by Default — No secrets in environment variables. No credentials in shell history. The CLI manages all sensitive data internally.
Eight Domains. Your Entire Stack.
Eight domains cover everything: core for content and capabilities, infra for services and databases, admin for users and agents, cloud for deployment, analytics for metrics, web for frontend, plugins for MCP servers, build for compilation.
This is the control plane architecture. Each domain maps to a capability boundary. Permissions scope to domains. Audit logs categorize by domain. Think in domains, operate in domains.
A2A protocol support means agents coordinate via systemprompt admin agents message. MCP tools execute via systemprompt plugins mcp tools. The CLI is the unified surface for your entire AI mesh.
- Domain Boundaries — Each domain (core, infra, admin, cloud, analytics, web, plugins, build) maps to a capability. Permissions scope to domains.
- Agent Coordination — A2A protocol via systemprompt admin agents message. Agents discover and invoke each other through the same CLI.
- MCP Integration — systemprompt plugins mcp tools lists and invokes MCP tools. The CLI is the control surface for your entire AI mesh.
Every Operation. Full Audit Trail.
Every command streams structured output to stdout. Every operation logs to the audit trail. Compose pipelines that transform data AND maintain complete operational visibility.
systemprompt infra logs audit --request-id abc123 traces any operation across services. See exactly what your AI infrastructure executed, when, and with what result.
Request inspection, tool execution logging, and agent message tracing are first-class operations. Debug production issues without leaving your terminal. The audit trail is always there.
- Pipeline Composition — Pipe JSON output through jq, xargs, and standard Unix tools. Build complex workflows from simple commands.
- Request Tracing — systemprompt infra logs audit --request-id abc123 traces any operation across all services with full context.
- Agent Message Tracing — Every A2A message, every tool invocation, every agent response logged with timestamps and trace IDs.
YAML Changes Deploy Instantly
YAML changes deploy instantly. Update an agent configuration, run systemprompt admin agents sync, and the change is live. Rebuild content with systemprompt core content publish. No code deploys for configuration changes.
Regenerate agents, restart services, republish content, sync MCP servers—all via CLI. CI/CD pipelines orchestrate with standard shell scripts. Idempotent operations mean safe retries.
Your AI superagent executes these same operations. Claude Code running your deploy script operates production with the same tooling you use. One harness for humans and agents alike.
- Zero-Code Deploys — YAML config changes deploy instantly. systemprompt admin agents sync applies changes without rebuilding binaries.
- Idempotent Operations — Run the same command twice, get the same result. Safe retries. CI/CD-friendly. No state corruption.
- Human-Agent Parity — Claude Code executes your deploy.sh with the same permissions and audit trail as you. One interface for everyone.
