Orion — for AI agents · Celestia network intelligence over MCP

Guide 01 — MCP

Connect your client

Hosted endpoint, no install, no keys — the server is strictly read-only. Pick your client below.

01

Claude Code one command

claude mcp add --transport http orion https://orionlabsone.cc/mcp

02

Cursor .cursor/mcp.json

{
  "mcpServers": {
    "orion": { "url": "https://orionlabsone.cc/mcp" }
  }
}

03

VS Code .vscode/mcp.json

{
  "servers": {
    "orion": { "type": "http", "url": "https://orionlabsone.cc/mcp" }
  }
}

04

Claude Desktop & stdio clients via mcp-remote

{
  "mcpServers": {
    "orion": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://orionlabsone.cc/mcp"]
    }
  }
}

05

Or clone the kit full source

git clone https://github.com/orioninfra/orion-agent-kit
cd orion-agent-kit && npm install && claude

Codex, Windsurf and Cline all speak the same stdio bridge config as Claude Desktop above — just swap the client name.

What the agent gets

Six analysis tools that return read verdicts rather than raw chain dumps, and every number keeps a provenance link.

Tool	Question it answers
`get_daily_brief`	What's happening on Celestia today?
`get_signals`	Any anomalies — today or on a given date?
`get_network_state`	Height, DA volume, validators, staking ratios, Nakamoto coefficient, TIA price
`list_reports`	What's in the daily archive?
`get_report`	Full report for one date
`search_reports`	When did the archive last mention "eclipse" or "block time"?

What a tool call actually returns

A real get_network_state response, pulled live right now:

loading live example…

Ask it things like

verdict"Any anomalies on Celestia today? Cite sources."

analysis"How concentrated is the validator set — what does the Nakamoto coefficient look like?"

comparison"Compare today's network state with last Monday's report."

archive search"When did Orion last flag block-time drift?"

Guide 02 — REST API

Call it directly, no MCP needed

The same data as plain JSON — no auth, no rate-limit headaches, ready for a cron job or your own backend. An OpenAPI 3.1 spec is available for function calling.

Endpoint	Returns
`GET /api/latest`	today's full report
`GET /api/reports`	archive index
`GET /api/reports/{date}`	report for a date
`GET /api/rss`	the daily brief as RSS

Try it now

curl https://orionlabsone.cc/api/latest

Returns the live snapshot and any active signals — this is the real response right now:

loading live example…

Minimal consumer, Python

No SDK, just requests — this is the same pattern Orion's own AI assistant writes when asked to build on Celestia data.

import requests

r = requests.get("https://orionlabsone.cc/api/latest", timeout=10)
report = r.json()

print(f"Height {report['snapshot']['height']:,} — "
      f"{len(report['signals'])} signal(s) today")
for s in report["signals"]:
    print(f"  [{s['severity']}] {s['title']}: {s['detail']}")

Guide 03 — AI chat

Talk to Orion directly

No setup at all — open orionlabsone.cc/celestia/aichat and ask. It's the same live Celestia data as the MCP tools and the REST API, grounded against the current snapshot so it won't invent a number it doesn't have.

Network analysis. Ask about signals, validator concentration, DAS mechanics, Blobstream lag, fibre readiness — answered from the live snapshot, not a static knowledge dump.

Writes real code. Pick a language and it stays in that mode for the rest of the conversation — submitting blobs, reading namespace data, building a validator alert bot, whatever you ask for next.

Connect MCP, from inside the chat. The "Connect MCP" button gives you the same endpoint as Guide 01 — so you can move from asking questions to wiring up your own agent without leaving the page.