seongsu-kang/tycono/getting-started

Tycono — How It Works

You are helping a user understand how to use the Tycono AI team orchestration plugin.

Core Concepts

Wave = A task you give to the AI team. Like a sprint. Agency = A team configuration (which roles, what domain knowledge). Role = A specialized AI agent (CTO, Engineer, QA, etc.)

Basic Flow

/tycono "your task"
  → CEO Supervisor analyzes task
  → Dispatches roles (CTO, Engineer, QA, etc.)
  → Each role works as a separate Claude session
  → Real-time notifications in your session
  → Results delivered

Checking Results

During the wave:

• Background notifications appear automatically (🔔 alerts)
• /tycono:tycono-status for current state

After the wave:

• Activity streams at .tycono/activity-streams/ses-*.jsonl
• Each file = one role's complete work log (thinking, tool calls, output)
• View with: cat .tycono/activity-streams/ses-cto-*.jsonl | python3 -m json.tool
• Or: grep '"type":"text"' .tycono/activity-streams/ses-critic-*.jsonl

For Existing Projects

If user has an existing project with code/knowledge:

/tycono:agency-create

This scans the project, suggests a team, and auto-generates + auto-verifies.

Common Questions

"How do I see what Critic said?" → cat .tycono/activity-streams/ses-critic-*.jsonl | grep '"type":"text"'

"How do I give a follow-up directive?" → Just run /tycono "follow-up task" — it creates a new wave in the same project context.

"The team did too much / too little" → Be more specific in the directive. Include what's already done, what the actual problem is, and what you expect.

"How do I set a default agency?" → Add to .tycono/config.json: { "defaultAgency": "your-agency-id" }

"How do I make a specific role think harder (or cheaper)?" → Add effort to that role's role.yaml (or an agency-wide default in agency.yaml):

# role.yaml
id: critic
model: claude-opus-4-6
effort: max     # low | medium | high | xhigh | max
 
# agency.yaml (applies to every role without its own `effort`)
default_effort: high

• Requires Claude CLI ≥ 2.1 (older CLIs don't accept --effort; server must be started with a newer CLI on PATH)
• Maps to Claude CLI --effort (API output_config.effort)
• Priority: role.yaml effort > agency.yaml default_effort > model default
• Tycono strips CLAUDE_CODE_EFFORT_LEVEL from the child env when a role has effort set — so a stray shell export doesn't silently override role settings
• max is Opus-class only (verified on opus-4-6 and opus-4-7; sonnet/haiku silently downgrade to high); the server emits a once-per-role warning when this happens
• Omit the field to use the model's default (≈ high for external users)
• Good picks: reasoning-heavy roles (Critic / Verdict-Judge) → max, low-stakes roles (log scribe / notifier) → low