Runtime Surface

Last updated: 2026-05-23

The important distinction is:

agenc-core

current agenc-core runtime

Public Launcher Install

Install the public launcher:

npm install -g @tetsuo-ai/agenc
agenc --help

Current launcher requirement:

Runtime development inside agenc-core currently requires:

Runtime Commands

Top-level runtime commands include:

agenc [options] [PROMPT]
agenc -p|--print [options] [PROMPT]
agenc help [command]
agenc init [--force]
agenc login
agenc logout
agenc whoami
agenc providers [--json] [--no-local-check]
agenc config <command> [args]
agenc plugin <command> [options]
agenc permissions <command>
agenc state export <agent-id>
agenc state import
agenc daemon start [--foreground]
agenc daemon stop
agenc daemon status
agenc daemon reload
agenc daemon restart
agenc agent start <objective>
agenc agent list
agenc agent attach <id>
agenc agent stop <id>
agenc agent logs <id>
agenc mcp <serve|add|list|get|remove|add-json|doctor>

Common flags:

-p, --print
--no-tui
--continue
--resume <session-id>
--profile <name>
--provider <name>
--model <id|provider:id>
--permission-mode <mode>
--autonomous
--proactive
--dangerously-bypass-approvals-and-sandbox
--yolo
--image <file|url|data-url>

Use agenc help <command> for command-specific help.

Daemon Runtime

The daemon is the local control plane. It owns:

Useful daemon commands:

agenc daemon status
agenc daemon start
agenc daemon start --foreground
agenc daemon reload
agenc daemon restart
agenc daemon stop

The launcher autostarts the daemon before invoking runtime commands unless AGENC_DAEMON_AUTOSTART=0 is set.

Runtime State

Runtime state is stored under AGENC_HOME when set, otherwise ~/.agenc.

The daemon uses the same home for:

Background Agents

Background agents are daemon-managed sessions that can run independently of the foreground TUI.

agenc agent start "fix the failing parser test"
agenc agent start --unattended-allow read,grep "audit imports"
agenc agent list
agenc agent attach <agent-id>
agenc agent logs <agent-id>
agenc agent stop <agent-id>

MCP, Plugins, Skills, And Tools

The runtime owns the generic coding-agent tool engine:

This is separate from Marketplace Agent Kit's marketplace-specific MCP and CLI surface.

Solana Transaction Guard

The runtime includes an opt-in SLM transaction guard for Solana-like tool calls. It runs at the tool-dispatch boundary and can block transaction-like actions before execution.

Enable with:

export AGENC_TRANSACTION_GUARD=slm
export AGENC_TRANSACTION_GUARD_MODEL=gemma4:e4b
export AGENC_TRANSACTION_GUARD_OLLAMA_URL=http://127.0.0.1:11434

The guard evaluates mutating Solana-like calls such as:

Read-only lookups such as solana balance, solana address, and solana config get are not evaluated.

Marketplace Boundary

Do not document agenc-core as the current marketplace canary gate.

Use Marketplace Agent Kit for:

See /docs/guides/marketplace-agent-kit and /docs/deployment/mainnet.