Agent — A configured AI assistant in Agent OS: id, name, model, personality, skills, tools, optional schedule. Stored in the DB or in ~/.agent-os/agents/.
Agent OS — Sulala Agent Operating System. Bun-based local platform: micro-agents, skills, task queue, scheduler, graphs, memory, channels. Default port 3010; home ~/.agent-os.
Channel — A messaging platform (Telegram, Slack, Discord, Signal, Viber). Agent OS receives messages via webhook (or polling), runs the default agent, and sends the reply back.
Config — ~/.agent-os/config.json. Holds API keys (OpenAI, OpenRouter, etc.) and channel credentials (tokens, default agent ids).
Conversation — A chat thread (conversation_id, messages). Used by the dashboard and channels for context; can be summarized.
Graph — A workflow of agents: nodes (each tied to an agent) and edges (from → to). Stored under ~/.agent-os/graphs/; run via API or dashboard.
Memory — Long-term storage (SQLite, optional vector search). API: write, search, delete. The memory skill exposes tools for the model.
Skill — An installable module under ~/.agent-os/skills/<name>/: skill.yaml (or SKILL.md) + optional tools.yaml, config.schema.json, SETUP.md. Registers tools and stores config in ~/.agent-os/configs/<skill_id>.json.
Task — A single run request for an agent. Lifecycle: queued → running → completed/failed. Stored in the DB.
Tool — A callable capability the model uses (e.g. echo, time, or tools from skills). Each has id, name, description, input schema. Central tool registry; agents get tools from their skills + built-in allowlist.
tools.yaml — Optional file in a skill folder. Defines tools (HTTP or exec) in a tools: array. Loaded by the skill loader and registered in the tool registry.
Webhook — Channel endpoint (e.g. POST /api/channels/telegram/webhook) that receives incoming messages from the platform. Config (token, default agent) is in config.
Workspace — Per-agent sandbox directory ~/.agent-os/workspaces/{agent_id}/. File tools are restricted to this path.
For more context, see Concepts.