Skip to main content
Connect your AI to Orbit in under two minutes. No infrastructure to set up, and no code required for most integrations.

​
Before you start

You’ll need:
  • An Orbit account at app.orbitagents.xyz
  • Your personal MCP endpoint URL — find it in Settings → setup inside your dashboard. It looks like this:
https://app.orbitagents.xyz/api/public/mcp/[your-key]
Keep this URL private — it’s your agent’s authentication key. The fastest way to start. Claude connects to Orbit via MCP (Model Context Protocol), the standard that lets AI tools share context and tools.

​
In Claude.ai or Desktop

  1. Open claude.ai → Settings → Integrations
  2. Click Add MCP Server
  3. Paste your Orbit MCP URL
  4. Name it “Orbit” and save
  5. Start a new conversation — Claude connects to Orbit automatically at the start of the session
  6. Copy the Skill and paste it in chat for Claude to save (important).

​
In Claude Code (terminal)

claude mcp add orbit https://app.orbitagents.xyz/api/public/mcp/[your-key]

​
Option 2 — Connect Codex or Cursor

Both support MCP servers. Add your Orbit MCP URL to the tool’s MCP config the same way as Claude Code. Your coding agent checks in with Orbit at the start of each session, reads your active decisions, and files its outputs to your Vault.

​
Option 3 — Perplexity

Support MCP servers. Add your Orbit MCP URL to the tool’s MCP config the same way as Claude Code. Your coding agent checks in with Orbit at the start of each session, reads your active decisions, and files its outputs to your Vault.

​
Option 4 — Connect via n8n

  1. In your n8n workflow, add an MCP Client node
  2. Set the server URL to your Orbit MCP endpoint
  3. Call orbit_heartbeat at the start of the workflow and orbit_result to file outputs
See the n8n connection guide for a full walkthrough.
More native MCP connections are coming in the future. Theoretically any agent can connect to Orbit with the MCP url and skill.

​
Your first test

After connecting, start a Claude session and ask:
“Check in with Orbit, tell me what decisions are active, and file a test result to the Vault.”
Claude will call orbit_heartbeat, receive your active decisions and knowledge, then call orbit_result. You’ll see both actions appear on your dashboard in real time.

​
What the agent does automatically

A properly connected agent will:
  • Call orbit_heartbeat at the start of every run — receiving your active decisions and approved knowledge
  • Call orbit_event throughout — so you see live progress, not just a final result
  • Call orbit_result when it finishes work — filing the output to your Vault with a trust level
  • Call orbit_heartbeat again at the end — closing the run cleanly
All of this happens automatically for agents using the Orbit MCP Skill. It can also:
  • Search your Vault before acting with orbit_search — so the agent finds what other agents already figured out instead of starting from scratch
  • File decisions with orbit_decide — turning a conclusion into a standing rule that future runs respect
  • Track structured data over time with orbit_table_write — pipelines, prospect lists, content queues, anything that needs to persist across runs
  • Attach files with orbit_file_init — PDFs, CSVs, images, exports that belong alongside the result
  • Reply to your messages with orbit_message — picking up instructions you left between runs and responding when it gets back online
  • Update its own behavior with orbit_update_self — when you give it a standing instruction (“always include a summary at the top”), it rewrites its own prompt so the change sticks forever, not just for the current session
  • Supersede or extend prior outputs — so a weekly report updates the same vault entry in place instead of creating a new one every time, keeping your Vault clean and your history intact
The full set of tools and what each one does is in the MCP Tools Reference.

Now what?

Three concrete first wins once your agent is connected.