Skip to main content
Glama
aniruddhaadak80

helio-commons

by aniruddhaadak80
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Remote

HELIO COMMONS

Map the next compute frontier.

Live app Next.js TypeScript License: MIT Feeds MCP

Live App · API health · MCP manifest · Issues

Helio Commons is a playable, evidence-first fieldbook for space data centers, Project Suncatcher, AI capability scenarios, and the long road toward superintelligence. It turns a research headline into a persisted flight plan, runs one explainable readiness engine, and leaves a replayable SHA-384 audit trail.

The demo is intentionally useful without an account or an API key: a user can create a mission, attach evidence, run a forecast, track a live research signal, export a brief, and let an MCP-compatible agent do the same through JSON-RPC.

Positioning: This is a planning aid for exploring assumptions. It is not a scientific forecast, investment recommendation, launch authorization, or safety certification.

✨ Features

  • Persistent mission CRUD — create, read, update, retire, and replay mission cards through Next.js route handlers.

  • Explainable readiness engine — evidence depth, horizon momentum, stage leverage, and coordination surface are itemized in every readout.

  • Live research desk — NASA Breaking News and arXiv AI feeds are normalized into one signal type, cached for 15 minutes, and backed by an attributed offline sample.

  • Agent interface — JSON-RPC initialize, tools/list, and tools/call, including mutating create_mission and update_mission tools.

  • Integrity story — every mutation appends SHA-384(previousSeal ‖ canonicalJson(event)); the verification route replays the chain.

  • Decision export — download a mission as a shareable Markdown brief with the current forecast and seal.

  • Fresh visual identity — a cobalt/amber launch-board schematic with animated telemetry, deliberately unlike a globe, dark-glass dashboard, or ticker template.

  • No-account core — the board is shared and public by design; auth is intentionally omitted because the first useful loop is research exploration, not private workspace ownership.

Related MCP server: operations-pulse

🗺️ Product map

Surface

What it does

Live API

/

Landing page and product thesis

—

/board

Create, list, and retire mission cards

GET/POST /api/missions

/mission/[id]

Edit a flight plan, attach evidence, replay forecast, verify audit, export brief

GET/PATCH/DELETE /api/missions/[id]

/signals

Read live/fallback research and attach a signal to a mission

GET /api/feed

/console

Discover and call MCP-style tools from the browser

POST /api/mcp

/method

Explain the engine and safety scope

—

/api/health

Storage mode, mission count, and audit health

GET /api/health

/api/forecast

Run the deterministic engine and return a forecast seal

POST /api/forecast

/api/audit/verify

Replay the full or mission-scoped chain

GET /api/audit/verify

/api/export

Download a mission brief

GET /api/export?missionId=…

🚀 Quickstart

git clone https://github.com/aniruddhaadak80/helio-commons.git
cd helio-commons
npm install
npm run dev

Open http://localhost:3000. The local experience needs no environment variables. Without DATABASE_URL, the app uses a seeded process-local store so the UI and API can be explored immediately. Production uses Neon Postgres; configure it with:

# Vercel production environment variable
DATABASE_URL=postgresql://...

The schema is in db/schema.sql. The app seeds its first three mission cards on an empty database and uses soft-delete tombstones so retired cards still participate in audit replay.

🧭 System architecture

flowchart LR
  Browser[Browser routes] --> API[Next.js route handlers]
  Agent[MCP client] --> JSON[JSON-RPC endpoint]
  JSON --> Domain[Shared domain functions]
  API --> Domain
  Domain --> Neon[(Neon Postgres)]
  Domain --> Engine[Forecast engine]
  Domain --> Audit[Audit seal writer]
  Domain --> Feeds[Feed normalizer]
  Feeds --> Fallback[Dated fallback sample]
  Neon --> Brief[Markdown exporter]

  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef agent fill:#34d399,color:#04060c,stroke:#34d399
  classDef caution fill:#fbbf24,color:#04060c,stroke:#fbbf24
  classDef infra fill:#94a3b8,color:#04060c,stroke:#94a3b8
  class Browser,API live
  class Domain,Engine,Audit,Brief engine
  class Agent,JSON agent
  class Feeds,Fallback caution
  class Neon infra

The important boundary is the shared domain layer: REST handlers, MCP tools, the UI actions, and exports do not maintain separate copies of the forecast or persistence rules.

📡 Live data layer

flowchart TB
  NASA[NASA Breaking News RSS] --> Parse[XML normalizer]
  Arxiv[arXiv AI Atom feed] --> Parse
  Parse --> Sort[Dedupe + sort + cap]
  Sort --> Live[Live signal response]
  Parse -. network failure .-> Sample[Attributed 2026 sample]
  Sample --> Fallback[Fallback response]
  Live --> Desk[Research desk]
  Fallback --> Desk
  Desk --> Mission[Attach evidence to mission]
  Mission --> Recalc[Recalculate forecast + seal]

  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef caution fill:#fbbf24,color:#04060c,stroke:#fbbf24
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef infra fill:#94a3b8,color:#04060c,stroke:#94a3b8
  class NASA,Arxiv,Parse,Sort,Live,Desk live
  class Sample,Fallback caution
  class Mission,Recalc engine

The feed route uses revalidate = 900. If both public feeds are unavailable, the app still returns dated signals from Google Research, Google DeepMind, GAO, AI 2027, and AI 2040 so the first paint and demo remain useful. The feed is research input, not an endorsement of any forecast.

🧮 Forecast engine

flowchart LR
  Input[Mission input] --> Normalize[Normalize evidence]
  Normalize --> E[Evidence depth 0..40]
  Normalize --> H[Horizon momentum 0..28]
  Input --> S[Stage leverage 0..24]
  Normalize --> C[Coordination surface 0..20]
  E --> Sum[Readiness index]
  H --> Sum
  S --> Sum
  C --> Sum
  Sum --> Confidence[Calibrated confidence]
  Confidence --> Seal[Forecast seal]
  Seal --> UI[UI factor readout]
  Seal --> API[REST / MCP response]

  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef agent fill:#34d399,color:#04060c,stroke:#34d399
  class Input,Normalize,E,H,S,C,Sum,Confidence,Seal engine
  class UI live
  class API agent

The same calculateForecast function serves the board, /api/forecast, and the MCP run_forecast tool. The current reference year is 2026; the factors are intentionally simple so a reader can challenge them.

  • Evidence depth: source count, verified count, and source-kind diversity.

  • Horizon momentum: how close the target year is to the reference year.

  • Stage leverage: watch, prototype, or deploy.

  • Coordination surface: diversity of evidence and presence of policy context.

  • Confidence: evidence quality adjusted for horizon distance and the user’s prior confidence.

🤖 Agent interface

flowchart LR
  subgraph Client["Agent client"]
    C["Coding agent"]
    M["MCP console"]
  end

  subgraph Service["Public Helio service"]
    J["JSON-RPC route"]
    D["Shared domain functions"]
    N[("Neon Postgres")]
  end

  C -- "initialize / tools/list" --> M
  M -- "JSON-RPC request" --> J
  J -- "validate arguments" --> D
  D -- "persist + SHA-384 seal" --> N
  N -- "mission + forecast" --> D
  D -- "structuredContent" --> J
  J -- "tool result" --> C
  C -- "tools/call create_mission" --> M
  C -- "tools/call verify_audit" --> M
  M -- "replay chain" --> D
  D -- "valid + checked count" --> C

  classDef agent fill:#34d399,color:#04060c,stroke:#34d399
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef infra fill:#94a3b8,color:#04060c,stroke:#94a3b8
  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  class C,M agent
  class J,D engine
  class N infra
  style Service fill:#22d3ee,color:#04060c,stroke:#22d3ee

Available tools:

Tool

Mutates?

Purpose

list_missions

no

Read active mission cards

create_mission

yes

Create and seal a mission

update_mission

yes

Patch fields/evidence and seal a revision

run_forecast

no

Run the shared engine and return a forecast seal

verify_audit

no

Replay the chain

export_mission

no

Render a Markdown brief

MCP client setup

{
  "mcpServers": {
    "helio-commons": {
      "type": "http",
      "url": "https://helio-commons.vercel.app/api/mcp"
    }
  }
}

The in-page console at /console proves the same endpoint with one-click initialize, tools/list, and a real mutating create_mission call.

🔐 Integrity and audit chain

flowchart TB
  Genesis[GENESIS] --> Create[created event]
  Create --> Update[updated event]
  Update --> Evidence[evidence revision]
  Evidence --> Retire[deleted tombstone]
  Create --> Hash[SHA-384 previousSeal + canonical JSON]
  Update --> Hash
  Evidence --> Hash
  Retire --> Hash
  Hash --> Verify[Replay endpoint]
  Verify --> Result{Valid chain?}
  Result -->|yes| Export[Export current seal]
  Result -->|no| Alert[Surface broken event]

  classDef verified fill:#34d399,color:#04060c,stroke:#34d399
  classDef risk fill:#fb7185,color:#04060c,stroke:#fb7185
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef caution fill:#fbbf24,color:#04060c,stroke:#fbbf24
  class Create,Update,Evidence,Retire,Hash,Verify,Export verified
  class Result,Alert risk
  class Genesis engine

The mission row stores the current prevSeal and seal; the audit table stores the full event data and monotonic sequence. GET /api/audit/verify recomputes the chain and reports the number of checked events.

🛰️ User journey

flowchart TB
  Idea[“Can orbital compute scale?”] --> Board[Create flight plan]
  Board --> Thesis[Write thesis + horizon]
  Thesis --> Source[Attach first source]
  Source --> Readout[Run readiness engine]
  Readout --> Gap{What is the gap?}
  Gap --> Evidence[Track another signal]
  Evidence --> Readout
  Gap --> Brief[Export decision brief]
  Brief --> Share[Share with a teammate or agent]
  Share --> Revise[Revise when reality changes]

  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef agent fill:#34d399,color:#04060c,stroke:#34d399
  classDef caution fill:#fbbf24,color:#04060c,stroke:#fbbf24
  class Idea,Thesis,Source,Brief,Revise live
  class Board,Readout,Gap,Evidence engine
  class Share agent

The three usefulness jobs are:

  1. A user can create/edit/delete a mission card so a research bet has an actionable home.

  2. A user can add evidence and run the deterministic engine so they can see which constraint deserves attention.

  3. A user can read live research, attach it, export a brief, or call the MCP tool so current research becomes a shareable next step.

🔌 API

Health

curl https://helio-commons.vercel.app/api/health

Create → read back

curl -X POST https://helio-commons.vercel.app/api/missions \
  -H 'content-type: application/json' \
  -d '{
    "name": "Optical fabric beats radio",
    "thesis": "Does short-range free-space optical linking unlock clustered orbital inference?",
    "category": "space-compute",
    "stage": "watch",
    "horizonYear": 2030,
    "confidence": 64,
    "evidence": []
  }'

Copy the returned mission.id, then read it back:

curl https://helio-commons.vercel.app/api/missions/<mission-id>

Update and forecast

curl -X PATCH https://helio-commons.vercel.app/api/missions/<mission-id> \
  -H 'content-type: application/json' \
  -d '{"stage":"prototype","evidence":[{"title":"A new test result","source":"Lab note","url":"https://example.com","kind":"primary","date":"2026-09-25","verified":true}]}'

curl -X POST https://helio-commons.vercel.app/api/forecast \
  -H 'content-type: application/json' \
  -d '{"missionId":"<mission-id>"}'

Verify and export

curl 'https://helio-commons.vercel.app/api/audit/verify?missionId=<mission-id>'
curl 'https://helio-commons.vercel.app/api/export?missionId=<mission-id>'

🚀 Deployment pipeline

flowchart LR
  Commit[Push to main] --> CI[GitHub Actions / Node 22]
  CI --> Lint[npm run lint]
  CI --> Build[npm run build]
  Lint --> Vercel[Vercel production build]
  Build --> Vercel
  Vercel --> Neon[(Neon DATABASE_URL)]
  Vercel --> Feeds[15-minute feed cache]
  Vercel --> Alias[Verified production alias]
  Alias --> Smoke[HTTP health + MCP + CRUD proof]

  classDef infra fill:#94a3b8,color:#04060c,stroke:#94a3b8
  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef verified fill:#34d399,color:#04060c,stroke:#34d399
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  class Commit,CI,Lint,Build,Vercel,Neon,Feeds,Alias infra
  class Smoke live
  class Lint,Build verified

🗺️ Roadmap

Now · Make the bet useful

  • Outcome: a user can create a mission and get an immediate, explainable readiness readout.

  • Outcome: evidence can be added, revised, and verified through a sealed chain.

  • Outcome: a live signal can become a persisted evidence item.

  • Outcome: an agent can discover and mutate the same domain through MCP.

flowchart LR
  Idea --> Mission --> Forecast --> Seal
  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef verified fill:#34d399,color:#04060c,stroke:#34d399
  class Idea,Mission live
  class Seal verified

Next · Make the board collaborative

  • Add lightweight signed identities or GitHub sign-in when a contributor needs ownership.

  • Add source snapshots and citation diffs so a claim can be compared across time.

  • Add a compare view for two missions and a “what changed” forecast delta.

  • Add a scheduled source-health view for feed freshness and fallback frequency.

flowchart TB
  Sources[Source snapshots] --> Diff[Citation diff]
  Missions[Two missions] --> Compare[Compare view]
  Diff --> Compare
  Compare --> Delta[Forecast delta]
  Delta --> Decide[Better next question]
  classDef live fill:#22d3ee,color:#04060c,stroke:#22d3ee
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef agent fill:#34d399,color:#04060c,stroke:#34d399
  class Sources,Missions,Compare live
  class Diff,Delta engine
  class Decide agent

Later · Open research protocol

  • Publish a signed event stream for independent replay.

  • Add optional user-supplied model adapters for qualitative synthesis while keeping the deterministic engine primary.

  • Add scenario packs for AI 2027 / AI 2040 with explicit assumptions and sensitivity ranges.

  • Submit the MCP endpoint to agent directories and publish reproducible research briefs.

flowchart LR
  Events[Signed events] --> Replay[Independent replay]
  Models[Optional model adapters] --> Brief[Scenario brief]
  Replay --> Trust[Trust layer]
  Brief --> Community[Community scenarios]
  Trust --> Community
  classDef verified fill:#34d399,color:#04060c,stroke:#34d399
  classDef engine fill:#a78bfa,color:#04060c,stroke:#a78bfa
  classDef caution fill:#fbbf24,color:#04060c,stroke:#fbbf24
  class Events,Replay,Trust verified
  class Models,Brief engine
  class Community caution

📚 Sources and attribution

Feed titles and summaries remain attributed to their publishers. Helio Commons does not claim that a source’s forecast is correct.

📁 Repository map

src/app/                 App Router pages, API routes, metadata, and loading/error states
src/components/          Nav, telemetry scene, board, detail, signals, and MCP console
src/lib/types.ts         Shared domain and normalized feed types
src/lib/engine.ts        Deterministic readiness engine
src/lib/db.ts            Neon + local persistence adapter and audit replay
src/lib/feed.ts          RSS/Atom normalization with dated fallback
src/lib/canonical.ts     Canonical JSON and SHA-384 sealing
src/lib/brief.ts         Markdown export renderer
src/lib/validation.ts    Shared REST/MCP input validation
db/schema.sql            Neon schema
public/mcp.json          Agent manifest
.github/workflows/ci.yml Node 22 lint/build workflow

⚠️ Safety and scope

This repository is adjacent to space infrastructure, AI capability forecasting, and governance. Do not use a Helio Commons readout to authorize a launch, certify hardware, make an investment decision, or predict a species-level future. The app is an explainable planning surface; real engineering, scientific review, legal review, and safety assessment remain human responsibilities.

🤝 Contributing

Read CONTRIBUTING.md, open an issue before large changes, and keep the deterministic engine and its factor readout covered by a reproducible example. See SECURITY.md before reporting a vulnerability.

📄 License

MIT © 2026 Aniruddha Adak. See LICENSE.

This server cannot be deployed

Pay for results, not tokens.Fund a bounty. Let AI agents compete and collaborate to solve your problems.agentbounties.appAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • Atlas MCP Server

    good2idnan
    F
    license
    Not graded
    quality
    D
    maintenance
    Enables AI agents and users to manage workspace files, monitor system metrics, take persistent notes, and retrieve weather data via MCP tools and resources.
    -

  • MCP Engineering Agent

    Lekha-2023
    F
    license
    Not graded
    quality
    B
    maintenance
    Enables a security-first engineering agent with an MCP-style JSON-RPC tool boundary, role-based authorization, and UTC audit events for every decision.
    -