Orion Agent Kit
MCP server providing Celestia network intelligence, including daily briefs, anomaly signals, staking and DA metrics, and live namespace and validator drill-downs, with each figure linked to its source.
README
<div align="center">
Orion Agent Kit
Celestia network intelligence for AI agents. An MCP server over Orion's daily analysis: anomaly signals, network verdicts, staking & DA metrics, live namespace and validator drill-downs — each figure linked back to the source it came from.
https://orionlabsone.cc/mcp
</div>
What this is
Orion watches Celestia mainnet continuously and publishes a daily machine-written brief: what moved, what drifted from baseline, and why it matters. This kit exposes that analysis to AI agents over the Model Context Protocol.
Think of it as an analysis layer for the Celestia agent ecosystem. It sits alongside knowledge bases (which answer "what is CIP-21?") and raw-data or transaction kits (which list blobs or build a delegate tx). Orion answers a different question: what is happening on the network right now, is it normal, and how did we get here?
Every number the server returns carries a provenance link to the Celenium or Celestia LCD endpoint it was read from, so the agent can cite its sources — and so can you. The server is read-only: no wallets, keys, or transactions anywhere in it.
Tools
Today & anomalies
| Tool | Question it answers |
|---|---|
get_daily_brief |
What's happening on Celestia today? (set include_narrative for Orion's written prose) |
get_signals |
Any anomalies (info / notable / critical) today or on date X? |
get_network_state |
Current height, DA volume, validators, staking ratios, Nakamoto coefficient, TIA price |
Trends & history
| Tool | Question it answers |
|---|---|
get_trends |
What changed since yesterday / last week? (day- and week-over-week deltas) |
get_metric_history |
Chart one metric (fees, blobs, price, staking ratio…) over N days |
list_reports |
What daily reports exist in the archive? |
get_report |
Full report for a given date |
search_reports |
When did the archive last mention "eclipse" or "block time"? |
Live drill-down (straight from Celenium)
| Tool | Question it answers |
|---|---|
get_top_namespaces |
Which rollups post the most data to Celestia right now? |
get_namespace |
Detail for one namespace: size, blob count, last activity |
get_top_validators |
Largest validators by stake — moniker, commission, jailed status |
Resources & prompts
Beyond tools, the server exposes MCP resources an agent can read or subscribe to:
orion://latest— the most recent daily reportorion://reports— the archive indexorion://report/{date}— any archived report by date
…and ready-made prompts to guide an agent:
daily_celestia_briefing— produce a grounded, source-cited daily briefingrollup_da_check— survey who dominates blobspace and the current DA cost context
What counts as a signal
Orion's detectors are event-driven, not state-driven — a persistently dominant namespace or a steady utilization day produces no signal, so what you see is what actually changed:
- DA utilization & fees — per-block volume and fees over the last day vs the recent daily baseline
- Blob-volume z-score — statistical spikes/drops in data posted
- Namespace shift — only when the DA leader changes or its share moves materially
- Staking & concentration — bonded-ratio moves, Nakamoto-coefficient changes, new jailings
- Market & protocol — notable TIA price moves, active governance proposals, new celestia-app releases
Each daily brief is written by an LLM narrator and passes an automated citation check — every number in the prose is verified against the structured data before publishing.
Quick start — hosted endpoint (no install)
The MCP server runs at https://orionlabsone.cc/mcp.
Claude Code
claude mcp add --transport http orion https://orionlabsone.cc/mcp
Cursor — add to .cursor/mcp.json (already included if you cloned this repo):
{ "mcpServers": { "orion": { "url": "https://orionlabsone.cc/mcp" } } }
VS Code — add to .vscode/mcp.json:
{ "servers": { "orion": { "type": "http", "url": "https://orionlabsone.cc/mcp" } } }
Claude Desktop / Codex / Windsurf / Cline / any stdio-only client — bridge with mcp-remote:
{
"mcpServers": {
"orion": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://orionlabsone.cc/mcp"]
}
}
}
Per-client walkthroughs: Claude Code · Claude Desktop · Cursor · Codex · VS Code
Quick start — clone & chat
git clone https://github.com/orioninfra/orion-agent-kit
cd orion-agent-kit
npm install
claude # Claude Code picks up .mcp.json automatically
Then just ask:
"Any anomalies on Celestia today?" · "What changed on Celestia since yesterday?" · "Which rollups post the most data right now?" · "How concentrated is the validator set?" · "When did the archive last flag block time?"
The bundled /brief and /signals commands wrap the two most common queries.
Run the server yourself
npm install
npm run start # Streamable HTTP on 127.0.0.1:8791/mcp
npm run start:stdio # stdio transport (for local MCP clients)
npm run smoke # end-to-end check of the tools against the live API
npm run typecheck # type-check the sources
Environment: PORT (default 8791), HOST (default 127.0.0.1), ORION_API_BASE (default https://orionlabsone.cc), CELENIUM_BASE (default https://api-mainnet.celenium.io/v1, used by the drill-down tools).
REST, without MCP
The underlying API is plain JSON, usable directly or via function calling with the OpenAPI spec:
| Endpoint | Returns |
|---|---|
GET /api/latest |
today's full report |
GET /api/reports |
archive index |
GET /api/reports/{date} |
full report for a date |
GET /api/rss |
the brief as an RSS feed |
How it fits together
AI agent (Claude / GPT / any MCP client)
│ MCP — read-only analysis tools, resources & prompts
▼
orion-agent-kit ──► Orion daily analysis (orionlabsone.cc/api)
│ │ every figure linked to its source
│ ▼
│ Celestia LCD · market data
└──────────► Celenium API (live namespace & validator drill-down)
License
Apache-2.0, Copyright 2026 Orion Labs. Questions, ideas, or a tool you wish existed: open an issue.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.