Advanced Features¶
Advanced SuperQode features for users who want more control over tools, safety, sessions, harnesses, and the TUI.
The Engine¶
-
Inside the Agent Loop
The run lifecycle and every guard: steering, auto-continue, reminders, doom-loop detection, deferred tools, compaction, rubric self-grading.
-
Tools Catalog
Every builtin tool: three edit dialects, interactive shell sessions, vision, peer agents, and the guarantees behind all of them.
-
Policies & Safety
The order of authority: hooks, exec-policy rules, env filtering, permission escalation, sandboxing.
-
Multi-Agent Workflows
Sub-agents, long-lived peer agents with live steering, A2A, and rubric quality gates.
-
Headless & CI
One-shot runs, JSON events, schema-validated output, rubric gates, session exports, worktree isolation.
User Controls¶
-
Safety & Permissions
Security model, permission rules, and dangerous operation handling.
Features¶
-
Memory & Learning
Persistent learning system that remembers patterns and learns from feedback.
-
Tools System
Comprehensive tool system with 20+ tools for AI coding agents.
-
Harness System
Reusable specs for runtime, model, tool, approval, event, and output policy.
-
Omnigent Compatibility
Import Omnigent
agent.yamlfiles into SuperQode harness specs while keeping SuperQode as the controlling harness. -
Session Management
Persistent session storage, sharing, and coordination.
-
Terminal UI
Rich terminal interface with widgets and interactive features.
How These Features Work Together¶
These advanced features help you run controlled coding-agent sessions:
- Memory & Learning: Learns from feedback
- Harness System: Defines reusable runtime, model, tool, approval, and event policy
- Session Management: Persists work
- Tools System: Provides capabilities
Configuration¶
Most advanced features are configured in superqode.yaml:
superqode:
permissions:
default: ask
session:
storage: file
memory:
enabled: true
Next Steps¶
- Configuration Reference - Full config options
- CLI Reference - Command reference