Feature guides

The reference pages say what every field and tab does. These guides say how the pieces fit together for one job at a time: what to turn on, in which order, and what to expect once it runs. Each one points back at the exact fields in agent.json5 and config.yaml.

• Agent-to-agent and the org chart: who may ask whom for work, and how the org chart and the allow-lists combine.
• Human in the loop and approvals: which actions wait for a person, and where the approval cards appear.
• Privacy and security: what an agent may read, reach and run, and how to tighten it.
• Multi-user: team accounts on one deployment, and per-customer isolation for a public-facing agent. Two different features.
• Subscriptions and monetization: charging for access to an agent.
• MCP servers: giving an agent external tools, with or without a sign-in.
• Webhooks: letting outside systems trigger an agent.
• Scheduler and heartbeat: recurring work, and the agent's own periodic check-in.
• Memory and self-learning: what an agent keeps between conversations and how it distils it.
• Notes: the agent's linked wiki.
• Issues: the shared tracker between people and agents.
• Revisions: the automatic history of configuration and markdown changes.
• Channels and connectors: WhatsApp, Telegram, Slack, Discord, email and the rest, including the OpenAI-compatible API.
• Cortex and Pulse: the background reasoning engine and the proactive suggestions it produces.
• Budgets: token ceilings and spending warnings.
• Ask the agent: self-configuration, magic links and the AgentFather admin tools.
• Beyond the deployment, A2A and MCP: talking to outside agents, and letting outside platforms reach yours.