Agent Features — Spectral Coding Agent by Aexol

Spectral Coding Agent

AgentFeaturesProblemsitsolves

End merge hell, code from anywhere, validate before you execute, ship with real design systems, and onboard into legacy codebases in minutes.

No more worktrees
Code from mobile
Predict execution
Design review
Conquer legacy
Project memory
Cross-Session Collaboration

AgentRadioShareknowledgeacrosssessions

Stop juggling git worktrees and painstaking merges. When multiple coding agents work on the same project folder, Agent Radio broadcasts critical observations — conventions, gotchas, architectural decisions — so every session stays aligned in real time without manual coordination.

No more worktrees — sessions share knowledge, not files
Painful merges eliminated — agents agree on conventions up front
Relevance-graded sharing (critical → low) with deduplication
Manual broadcast for must-know findings across sessions
Agent Radio4 sessions online

The parser package must remain zero-dependency.

criticalSession #4 — Mac Studio · 2 min ago
a1b2c3d4e5f6

Generated files (models.ts, zeus/) are NEVER hand-edited.

highSession #2 — Dev Server · 8 min ago
b2c3d4e5f6a1

All Deno imports must use .ts extensions.

highSession #1 — Laptop · 14 min ago
c3d4e5f6a1b2

The landing/ directory uses npm, NOT deno.

mediumSession #3 — Workstation · 22 min ago
d4e5f6a1b2c3

The landing fonts are loaded via next/font in the root layout.

lowSession #2 — Dev Server · 45 min ago
e5f6a1b2c3d4
Mac Studioconnected
mac-studio.local·2 sessions·Claude 3.5 Sonnet
Dev Serverconnected
dev-box.internal·1 sessions·Claude 3 Opus
Laptopconnected
thinkpad.local·1 sessions·GPT-4o
Workstationoffline
ws-4.corp·0 sessions·Claude 3.5 Sonnet
3 of 4 machines online · 4 active sessions
Multi-Machine Fleet

Codefromanywhere,onanymachineyouown

Code even from mobile by connecting to your running machines — or share them with your entire team. Switch between Mac Studio, dev server, laptop, or workstation from a single UI. Your infrastructure is always a click away, no matter where you are.

Code from mobile — connect to any machine in your fleet
Team-wide access — share machines with your whole team
One-click switch — session context resets cleanly
Real-time status for every machine with refresh on demand
Prompt Quality Gate

Validatebeforeyouexecute

Cut bad decisions before they happen. The Prompt Quality Gate evaluates your prompt against the project’s entire memory context — observations, reflections, AGENTS.md conventions — and shows you the predicted execution path in seconds. See the score, fix inconsistencies, and only hit send when you’re sure.

See the predicted execution path before you commit
Eliminates inconsistent prompts that waste iterations
1-100 quality score with traffic-light gradient
3-5 refined alternatives when score is below threshold
Designer Extensions

Designsystems&skills,builtintotheagent

Stop guessing on design. The agent ships with 150+ curated design systems and structured skill workflows — pick a system, review the output against brand tokens, and ship polished HTML/CSS on the first pass. The agent runs the design review so you can focus on delivering.

150+ Design SystemsClick to preview
Apple
Tech · CSS tokens · Typography · Spacing
HeadingsBody textButtonsCardsInputs
Design SkillsStructured workflows
Auto-Research
Auto-Research

Theagentstudiesyourprojectforyou

Drop into a legacy project with confidence. Auto-Research scans the codebase’s structure, dependencies, and conventions to generate agent extensions tailored to the project. Unknown territory becomes supercharged territory — the agent learns the project before you write a single line.

Enter legacy codebases — extensions generated automatically
Supercharge development in unknown project territory
4-phase pipeline: collect → analyze → generate → validate
Extensions auto-register and persist across sessions
Observational Memory

Persistentprojectknowledge,acrosseverysession

The agent maintains a durable memory of project conventions, architectural decisions, and discovered patterns. Observations persist across sessions and are searchable by every agent that connects to the project.

AUTO

Automatic Observation

The agent records conventions and patterns as it works. No manual documentation needed — every discovery becomes durable project knowledge.

PERSISTENT

Cross-Session Recall

New sessions hydrate from the project observation store. The agent knows the codebase conventions before it writes a single line.

SEARCHABLE

Semantic Search

Search past observations by convention keywords, file paths, or technology names. Find relevant context instantly.

PRIORITY

Relevance Grading

Every observation is graded critical/high/medium/low. The agent prioritizes critical facts and never forgets NEVER-EDIT rules.

AUTO

Memory Compaction

Automatic context compaction keeps the agent's working memory lean. Old observations are pruned while preserving critical knowledge.

VISIBLE

Memory Status

Live status dashboard showing observation count, compaction phase, token usage, and reflection quality. Full transparency into agent memory.

Complete Tool Suite

Everytooltheagentneeds,builtin

From autonomous loops to browser automation, Kanban integration to MCP server management — the agent ships with everything needed for production AI-native development.

Workflow

Autonomous Loop

Let the agent iterate until satisfied — set max iterations and a goal. The agent self-corrects and reports when done.

Prompt Queue

Enqueue follow-up prompts during streaming. The agent processes them in order without blocking the UI.

Fork & Compact

Fork a conversation into a compacted session. Keeps context lean while preserving the full history for reference.

Infrastructure

MCP Servers

Manage external tool servers per team — connect, test, enable/disable. Compatible with the Model Context Protocol.

Native Extensions

Browse and manage extensions loaded on the machine. Designer tools, auto-research, Kanban bridge, and more.

Browser Tools

Full browser automation — navigate, click, type, screenshot, evaluate JS. Headless Chrome via Playwright.

Session

File Search (@)

Type @ in the composer to search project files. Auto-completes relative paths from the project's file tree.

Kanban Integration

Pick tasks from the Studio Kanban board. Agent can move tasks through columns as it works.

Input History

Up/Down arrows cycle through previous prompts. Terminal-style navigation keeps you in flow.

Ready to start

Try the agent today

Download the CLI, start a server, and connect your browser. All features are available with zero configuration — the agent discovers your project on its own.