Super Multica

A multi-component architecture for distributed agent systems.

Project Structure

src/
├── agent/              # Core agent module
│   ├── context-window/ # Token-aware context management
│   ├── profile/        # Agent profile management
│   ├── session/        # Session persistence with compaction
│   ├── skills/         # Modular skill system
│   └── tools/          # Agent tools
│       └── web/        # Web fetch and search tools
├── gateway/            # WebSocket gateway for distributed communication
├── hub/                # Multi-agent coordination hub
├── client/             # Client library
├── console/            # NestJS console application
└── shared/             # Shared types and gateway SDK
    └── gateway-sdk/    # Gateway client SDK

apps/
└── web/                # Next.js web application

packages/
└── sdk/                # SDK package for external use

skills/                 # Bundled skills (commit, code-review)

Getting Started

pnpm install

Environment Configuration

The Agent requires LLM provider credentials. Copy the example and fill in your values:

cp .env.example .env
# Edit .env with your API keys

Example .env for OpenAI:

export LLM_PROVIDER=openai
export OPENAI_API_KEY=sk-xxx
export OPENAI_BASE_URL=https://api.openai.com/v1
export OPENAI_MODEL=gpt-4o

Load the environment before starting services that use the Agent:

# Hub Console (requires LLM env vars)
source .env && pnpm dev:console

# Agent CLI
source .env && pnpm agent:cli "hello"

# Gateway (no LLM env vars needed)
pnpm dev:gateway

See .env.example for all supported providers (OpenAI, Anthropic, DeepSeek, Kimi, Groq, Mistral, etc.).

Configuration Priority

Each setting is resolved in order (first match wins):

1. CLI argument — --provider, --model, --api-key, --base-url
2. Environment variable — LLM_PROVIDER, OPENAI_MODEL, OPENAI_API_KEY, OPENAI_BASE_URL, etc.
3. Session metadata — restored from previous session
4. Default — kimi-coding provider with kimi-k2-thinking model

Agent CLI

Use the agent module directly from the CLI for isolated testing.

# New sessions get a UUIDv7 ID (shown on start)
pnpm agent:cli "hello"
# [session: 019c0b0a-b111-765c-8bbd-f4149beac9c4]

# Continue a session
pnpm agent:cli --session 019c0b0a-b111-765c-8bbd-f4149beac9c4 "what did I say?"

# Or use a custom session name
pnpm agent:cli --session demo "remember my name is Alice"
pnpm agent:cli --session demo "what's my name?"

# Override provider/model
pnpm agent:cli --provider openai --model gpt-4o-mini "hi"

# Use an agent profile
pnpm agent:cli --profile my-agent "hello"

# Set thinking level
pnpm agent:cli --thinking high "solve this complex problem"

Sessions

Sessions persist conversation history to ~/.super-multica/sessions/<id>/. Each session includes:

• session.jsonl - Message history in JSONL format
• meta.json - Session metadata (provider, model, thinking level)

Sessions use UUIDv7 for IDs by default, providing time-ordered unique identifiers.

Context Window Management

The agent automatically manages context windows to prevent token overflow:

• Token-aware compaction - Tracks token usage and compacts when approaching limits
• Compaction modes: tokens (default), count (legacy), summary (LLM-generated)
• Configurable safety margins - Ensures space for responses
• Minimum message preservation - Keeps recent context intact

Agent Profiles

Agent profiles define identity, personality, tools, and memory for an agent. Profiles are stored as markdown files in ~/.super-multica/agent-profiles/<id>/.

Profile CLI

# Create a new profile with default templates
pnpm agent:profile new my-agent

# List all profiles
pnpm agent:profile list

# Show profile contents
pnpm agent:profile show my-agent

# Open profile directory in file manager
pnpm agent:profile edit my-agent

Profile Structure

Each profile contains:

• identity.md - Agent name and role
• soul.md - Personality and behavioral constraints
• tools.md - Tool usage instructions
• memory.md - Persistent knowledge
• bootstrap.md - Initial conversation context

Skills

Skills are modular capabilities that extend agent functionality. They are defined in SKILL.md files with YAML frontmatter.

Built-in Skills

Located in /skills/:

• commit - Git commit helper following conventional commits
• code-review - Code review assistance

Skill Format

---
name: Skill Name
description: What it does
version: 1.0.0
metadata:
  emoji: 📝
  requiresBinaries: [git]
  platforms: [darwin, linux]
  tags: [git, tools]
---

## Instructions
(markdown instructions for the agent)

Eligibility Filtering

Skills can specify requirements:

• requiresBinaries - Required CLI tools
• requiresEnvVars - Required environment variables
• platforms - Supported platforms (darwin, linux, win32)

Skills are automatically filtered based on the current environment.

Agent Tools

exec

Execute short-lived shell commands and return output. Commands running longer than the timeout are automatically backgrounded.

exec({ command: "ls -la", cwd: "/path/to/dir", timeoutMs: 30000 })

process

Manage long-running background processes (servers, watchers, daemons). Output is buffered (up to 64KB) and terminated processes are automatically cleaned up after 1 hour.

# Start a background process (returns immediately with process ID)
process({ action: "start", command: "npm run dev" })

# Check process status
process({ action: "status", id: "<process-id>" })

# Read process output
process({ action: "output", id: "<process-id>" })

# Stop a process
process({ action: "stop", id: "<process-id>" })

# Clean up terminated processes
process({ action: "cleanup" })

glob

Pattern-based file discovery using fast-glob.

glob({ pattern: "**/*.ts", cwd: "/path/to/dir" })

web_fetch

Fetch and extract content from URLs with intelligent content extraction.

# Basic fetch (returns markdown)
web_fetch({ url: "https://example.com" })

# With options
web_fetch({
  url: "https://example.com",
  outputFormat: "markdown",  # or "text"
  extractor: "readability"   # or "turndown" for full page
})

Features: SSRF protection, response caching, max 50KB output.

web_search

Search the web using Brave or Perplexity AI.

# Basic search
web_search({ query: "typescript best practices" })

# With provider options
web_search({
  query: "latest AI news",
  provider: "brave",     # or "perplexity"
  count: 5,
  freshness: "pw"        # past week (Brave: pd/pw/pm/py)
})

Distributed Architecture

Gateway

The WebSocket gateway enables distributed multi-agent communication:

• Real-time message passing between agents
• Streaming support for long-running operations
• RPC-style request/response patterns

Hub

The Hub manages multiple agents and gateway connections:

• Agent lifecycle management
• Communication channel coordination
• Device identification and tracking

Scripts

Agent Commands

• pnpm agent:cli - Run the agent CLI for module-level testing
• pnpm agent:interactive - Interactive REPL mode
• pnpm agent:profile - Manage agent profiles

Development

• pnpm dev - Run full stack in development mode
• pnpm dev:gateway - Run gateway only
• pnpm dev:console - Run console only
• pnpm dev:web - Run web app only

Build & Test

• pnpm build - Build for production
• pnpm build:sdk - Build SDK package
• pnpm start - Run production build
• pnpm typecheck - Type check without emitting