mcp-ocei-oracle

README.md

# OCEI™ v2.2 — Orbital Compute Efficiency Index & Arbitrage Oracle (MCP Server) - OCEI Index

[![OCEI v2.2 Live Benchmark](https://orbitcompute.ai/v1/index/ocei/badge.svg)](https://orbitcompute.ai/ocei)

Official **Model Context Protocol (MCP)** server and deterministic reference client for **[OrbitCompute.ai](https://orbitcompute.ai)**, publisher of the **OCEI™ v2.2 (Orbital Compute Efficiency Index)**.

Provides cryptographically attested (**RFC 8785 JSON Canonicalization + ECDSA P-256 `ES256`**) real-time evaluations comparing orbital AI compute shells (**650 km Dawn-Dusk SSO**, **320 km VLEO Edge**, **550 km LEO**, **1,500 km MEO**) against **Terrestrial Tier-IV Utility Grids** using 6 live **NOAA SWPC & GOES satellite telemetry streams**.

## Live Specification & Endpoints

* **Interactive Control Room & Multi-Physics Simulator:** [https://orbitcompute.ai](https://orbitcompute.ai)
* **Formal Mathematical Specification & CODATA Registry:** [https://orbitcompute.ai/ocei](https://orbitcompute.ai/ocei)
* **Token-Optimized Markdown Spec (for LLMs):** [https://orbitcompute.ai/ocei.md](https://orbitcompute.ai/ocei.md)
* **Live Attested JSON Feed:** [https://orbitcompute.ai/v1/index/ocei](https://orbitcompute.ai/v1/index/ocei)
* **Ultra-Compact M2M Feed (<160 Tokens):** [https://orbitcompute.ai/v1/index/ocei?format=compact](https://orbitcompute.ai/v1/index/ocei?format=compact)
* **Global Verification Public Key (JWKS):** [https://orbitcompute.ai/.well-known/jwks.json](https://orbitcompute.ai/.well-known/jwks.json)
* **OpenAPI 3.1.0 Specification:** [https://orbitcompute.ai/v1/openapi.json](https://orbitcompute.ai/v1/openapi.json)

## Exposed MCP Tools

1. **`evaluate_orbital_compute_regime`**: Computes live OCEI-Spot, OCEI-Structural, and 60m-TWAP scores, L1 solar-wind shockwave lead time (+25 to +62 min), solar beta-angle ($\beta$) eclipse battery mass, 2-sided Stefan-Boltzmann radiator area with Earth IR/albedo view-factor absorption, Silicon $T_j$ thermal throttling, 25-MW fault-domain Young-Daly SEU checkpoint intervals, and WACC-adjusted $/1M Token unit economics.
2. **`compute_arbitrage_dispatch`**: Evaluates real-time terrestrial electricity spot prices ($/MWh) against the orbital breakeven threshold to output an attested workload routing signal (`ORBIT_PRIMARY`, `HYBRID_SPLIT`, or `TERRESTRIAL_PRIMARY`).
3. **`solve_optimal_workload_allocation`**: Deterministic constraint solver returning the optimal percentage split across `SSO-650`, `VLEO-320`, and `Terrestrial-Spot` based on workload class and latency tolerance.

## Quickstart (Claude Desktop / Cursor / Windsurf / MCP Clients)

### Option A: Direct Streamable HTTP / Remote MCP
```json
{
  "mcpServers": {
    "ocei-oracle": {
      "url": "[https://orbitcompute.ai/mcp](https://orbitcompute.ai/mcp)"
    }
  }
}

Deploy Server

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

B3.3/5.0

Scored 2026-11-12 23:22 across 3 tools

Disambiguation3/5

The three tools cover related but overlapping decisions: evaluate_orbital_compute_regime already assesses orbital vs. terrestrial viability, while compute_arbitrage_dispatch and solve_optimal_workload_allocation both output routing/allocation decisions (the former a coarse 3-way directive, the latter a percentage split). An agent could plausibly reach for either of the latter two, or start with the evaluator, without a clear rule for ordering. Descriptions help, but boundaries between 'dispatch' and 'allocate' remain fuzzy.

Naming Consistency4/5

All three names are snake_case and verb-led (evaluate_, compute_, solve_), giving a predictable pattern. Minor deviation: compute_arbitrage_dispatch reads awkwardly (verb + noun-phrase 'arbitrage_dispatch') compared to the cleaner verb_object form of the other two, but the convention is still recognizably consistent.

Tool Count4/5

Three tools is tightly scoped to a narrow analytical domain (orbital compute economics), and each maps to a distinct decision stage: evaluation, real-time arbitrage, and allocation optimization. It is on the lean side for the breadth of physics and economics mentioned, but nothing feels gratuitous.

Completeness4/5

The surface covers the core lifecycle — assess the regime, decide routing from spot prices, and solve for an optimal split — with no obvious dead ends for the stated purpose. Gaps are minor: there is no tool to inspect or override model constants/thresholds, nor a what-if/sensitivity or report-explanation operation, which an agent may want after getting scores.

Maintenance

ActivityMaintained

ResponsivenessNo recent community issues

TDQS

Maintenance

Resources