Skip to content

First Run

After completing Installation and starting the stack, here's what to expect.

What Happens on First Boot

  1. Database migrations — creates all tables (settings, workflows, resources, knowledge, memory, etc.)
  2. Config seeding — environment variables are written to the settings table
  3. Default LLM providers — Gemini, Claude, Ollama, and LM Studio providers are created (disabled if no API key set)
  4. Default workflows — example workflows are seeded (Tech News Digest, Web Research, etc.)
  5. Default RSS feeds — a set of news and tech feeds are created as resources
  6. Self-awareness indexing — architecture docs are chunked and stored in the knowledge base

Access the Admin UI

Open http://localhost:5001 in your browser and log in with the ADMIN_PASSWORD you set in .env.

The dashboard shows:

  • Active workflows and recent runs
  • LLM provider status and usage
  • Skill registry (core + dynamic)
  • System health

Test Telegram

Send /ping to your bot — you should receive pong back.

Try a few more commands:

/status          → System stats (uptime, memory, skills)
/skills          → List all registered skills
/workflows       → List workflows with status
/help            → Full command reference

Run Your First Workflow

  1. Go to Admin > Workflows
  2. Find "Tech News Digest" and click Run Now
  3. Watch the execution progress in real-time
  4. The result is delivered to your Telegram chat

Or trigger it from Telegram:

/run Tech News Digest

/search what is the BEAM virtual machine?

This invokes the web_search skill — searches DuckDuckGo, fetches top results, and synthesizes an answer via your configured LLM.

Configure MCP (Optional)

If you want to connect AI clients (Claude Code, Cursor):

  1. Go to Admin > Config
  2. Set mcp.api_key to a strong random value (category: mcp, sensitive: true)
  3. Configure your client — see MCP Client Setup

Next Steps