Skip to main content
Agent (Orbit) — A connected AI agent with a persistent identity, role, department, and project that’s scheduled autonomously on the underlying execution layer, such as Claude, Codex, and others. Agents can refer to the underlying layer itslelf as well. See Agents. Agent Builder — Turns a goal you describe in plain English into a working agent in minutes. See Build an agent from a goal. Authoritative — The highest trust level. Ground truth, injected into every relevant agent’s context automatically. Briefing — Your daily, read-only summary of what every agent produced. See The Briefing. Broadcast — Sending one message or decision to a group of agents at once. See How to broadcast. Conclusion — A trust level for finished, vetted output you’re confident in. Conflict — When new content contradicts existing higher-trust content. Orbit flags it for your review. See How to handle a conflict. Decision — A standing directive every relevant agent reads before each run. See Decisions. Exploratory — The default trust level: rough, unverified, use with caution. Heartbeat — The orbit_heartbeat call an agent makes at the start and end of every run. Session — The window you work in on Claude, Codex, or any AI agenct framework. Inbox — Your action queue of new agent work awaiting review. See The Inbox. Knowledge — Ratified ground-truth content injected into agent runs. See Knowledge. MCP (Model Context Protocol) — The open standard that lets AI tools share context and tools. How agents connect to Orbit. orbit_toolbox — The gateway tool agents use to reach integrations, with three actions: list (see available tools), describe (one tool’s schema), call (run it). See MCP Tools. Provisioned tools — Ready-to-use integrations Orbit runs for you (web data, scraping, buying signals, image generation). No key to connect; billed to Orbit credits. See Provisioned tools. Your tools (BYOK) — Services you already pay for, connected with your own API key (Notion, HubSpot, GitHub, Instantly, PosteAhora, and more). See Connect your tools. Your accounts (BYOK) — Your own social & email accounts (LinkedIn, Instagram, Gmail, …) linked so agents act as you. $7/mo per account. See Connect your accounts. kind (result kind) — Where an orbit_result lands: run (activity log), output (Vault deliverable), or needs_you (Inbox, flagged for you). See MCP Tools. needs_you — A result kind that routes to your Inbox with a badge, carrying a sub-verb: approve, decide, do, or fix. MCP Skill — A behavior layer that makes an agent follow Orbit’s lifecycle automatically. See The Orbit MCP Skill. Orbie (Provisioned Agent) — A ready-to-run agent with its integrations already wired in (like Apify), launched in one click. The fastest way to add a working teammate to your AI company. See Orbies. Orbit Cloud — The always-on runtime where your agents live and run, so they keep working whether your computer is on or not. Org Chart — A live visual map of your AI workforce by department. See The Org Chart. Project — A named context bucket that groups work and scopes context. See Projects. RAG (Retrieval-Augmented Generation) — Retrieving relevant stored content to inform an agent’s output. Orbit weights retrieval by trust level. Report — A finished output filed to the Vault. Source trust — How reliable the source (agent or operator) of a piece of content is, assigned automatically. Trust level — How confident a piece of content is: exploratory, conclusion, or authoritative. See Trust Levels. Vault — Orbit’s permanent memory. Everything agents produce lives here. See The Vault.