https://github.com/mapleleaflatte03/meridian-intelligence
Meridian Intelligence — Autonomous AI intelligence service with 7 specialized agents, daily briefs, and USDC payments via x402 protocol
https://github.com/mapleleaflatte03/meridian-intelligence
ai-agents ai-intelligence artificial-intelligence autonomous-agents base-l2 daily-brief llm mcp mcp-server model-context-protocol openclaw usdc x402
Last synced: about 2 months ago
JSON representation
Meridian Intelligence — Autonomous AI intelligence service with 7 specialized agents, daily briefs, and USDC payments via x402 protocol
- Host: GitHub
- URL: https://github.com/mapleleaflatte03/meridian-intelligence
- Owner: mapleleaflatte03
- Created: 2026-03-16T04:55:36.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-03-31T06:15:06.000Z (2 months ago)
- Last Synced: 2026-03-31T08:36:47.913Z (2 months ago)
- Topics: ai-agents, ai-intelligence, artificial-intelligence, autonomous-agents, base-l2, daily-brief, llm, mcp, mcp-server, model-context-protocol, openclaw, usdc, x402
- Language: Python
- Size: 2.27 MB
- Stars: 8
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Meridian — Loom-first governed agent runtime portal
Loom is the flagship local runtime product, Kernel is the runtime-neutral governance core, and the Meridian apps are first-party proof workloads.
The live host keeps that hierarchy explicit instead of collapsing it into one vague category claim.
Home ·
Loom ·
Why ·
Compare ·
Demo ·
Pilot ·
Support ·
Boundary
# Meridian Intelligence
This repo is the public Meridian portal plus the first-party Meridian apps that run on Loom + Kernel. It is not the governance kernel, and it is not the runtime itself. The product front door is now Loom.
## Launch Positioning
Meridian currently presents a Loom-first hierarchy around four ideas:
- Meridian Loom as the flagship local runtime product.
- Meridian Kernel as the runtime-neutral governance core.
- Meridian Intelligence plus bounded Trust Ops as first-party workloads that prove the runtime under real pressure.
- A truthful host boundary that separates verified live routes from future deployment claims.
## 1-Command Install
If the operator question is "what do I install?", the answer is Loom:
```bash
curl -fsSL https://raw.githubusercontent.com/mapleleaflatte03/meridian-loom/main/scripts/install.sh | bash
```
The installer story belongs to Meridian Loom. This repo links to that runtime surface while keeping the live portal and host boundary honest.
## Product Hierarchy
The current Meridian story across the stack is:
- **Loom runtime:** the product front door and official first-party runtime.
- **Kernel boundary:** five runtime-neutral governance primitives — Institution, Agent, Authority, Treasury, Court.
- **First-party Meridian apps:** Intelligence and bounded Trust Ops workloads that prove what Loom + Kernel can carry.
Loom carries execution, sessions, channels, skills, personal agents, and memory/context. Kernel carries authority, treasury, court, warrants, sanctions, and the runtime contract. Commitment remains a Meridian platform primitive composed above Kernel, but it is not the category center or the runtime front door.
## Three-Part Architecture
- [meridian-loom](https://github.com/mapleleaflatte03/meridian-loom): official first-party local runtime, install path, CLI, personal agents, channels, memory, and proof receipts.
- [meridian-kernel](https://github.com/mapleleaflatte03/meridian-kernel): governance truth, institution, authority, treasury, court, and the runtime contract Loom consumes.
- [meridian-intelligence](https://github.com/mapleleaflatte03/meridian-intelligence): portal, live host surface, first-party workflows, subscriptions, and bounded customer-facing routes.
## What Is Live Today
- Loom as both the live execution runtime on this host and the installable local runtime.
- The public web portal, Loom landing page, compare page, demo page, and boundary note.
- Competitor intelligence as the first Loom-backed first-party workflow.
- Bounded Trust Ops primitives as additional first-party workloads on the same host.
- A bounded live-host stance: Meridian-facing execution routes on this host now fail closed instead of silently falling back to older runtime paths.
- Meridian platform surfaces that expose the sixth primitive, Commitment, above the kernel boundary.
- A public runtime comparison page that explains where Loom is sharper than OpenClaw, OpenFang, IronClaw, TEMM1E, Goose, OpenHands, CrewAI, and LangGraph and where it still has to earn the right to win: https://app.welliam.codes/compare
## What Is Not Claimed
- Full hosted runtime replacement.
- Every possible deployment route running live on this host.
- Universal live automation without treasury, policy, and boundary proof.
## Intelligence Workflow (Current Vertical)
Competitive intelligence is the first managed first-party workflow vertical. What exists today is the workflow, the runtime, and the product surface. What is intentionally still narrow is the delivery promise. The public demo and battlecard pages include qualified reference examples; they are not meant to imply that every illustrated output is a currently live customer artifact on the public host.
- **Cited competitor alerts** — findings on pricing changes, launches, API updates, and deprecations
- **Curated intelligence briefs** — top competitive moves with action items
- **Battlecards on demand** — structured competitor snapshots for sales enablement
- **Competitor watchlists** — track specific companies through the governed workflow
### Current Entry Path
Start with the demo and pilot surfaces:
- Loom: https://app.welliam.codes/loom
- Demo: https://app.welliam.codes/demo
- Pilot: https://app.welliam.codes/pilot
- Support the work: https://app.welliam.codes/support
- Contact: Telegram [@Enhanhsj](https://t.me/Enhanhsj) or email `nguyensimon186@gmail.com`
The honest current offer is a founder-led paid pilot plus an evidence-backed checkout capture rail. More cash in treasury only clears the reserve gate; broader automated delivery still waits for customer-backed phase progression, runtime preflight, and constitutional approval where required.
If you want to back Meridian without pretending that support equals customer delivery, use the dedicated support path instead of the pilot flow.
If you are confused about support vs pilot vs customer revenue vs future contributor payouts, read:
- [Meridian Doctrine](company/MERIDIAN_DOCTRINE.md)
---
## MCP Tools
Connect via SSE: `https://app.welliam.codes/sse`
```json
{
"mcpServers": {
"meridian": {
"url": "https://app.welliam.codes/sse"
}
}
}
```
| Tool | Price | Description |
|------|-------|-------------|
| `intelligence_latest_brief` | **$0.50 USDC** | Metered brief tool for the founding Meridian institution; external-customer settlement proof is still pending |
| `intelligence_on_demand_research` | **$2.00 USDC** | Metered research tool with sourced findings on the current public MCP surface |
| `intelligence_competitor_snapshot` | **$3.00 USDC** | Metered competitor snapshot tool; public battlecard layouts remain reference examples |
| `intelligence_qa_verify` | **$1.00 USDC** | Metered QA verification of claims or text |
| `intelligence_weekly_digest` | **$1.50 USDC** | Metered digest tool; public weekly brief layouts remain reference examples |
| `company_info` | **FREE** | Meridian capabilities and pricing |
On the live host today, every MCP tool call is audited and metered for the
founding Meridian institution only. The shared runtime-core taxonomy classifies
that path as `mcp_service` with identity model `x402_payment` and scope
`founding_service_only`. Multi-institution MCP routing is not live.
The live agent registry also surfaces `runtime_binding` on each governed agent
record, and that same field is visible through `GET /api/agents` and the
`agents` array plus the top-level `agent_runtime_bindings` tranche inside
`/api/status`. That makes the agent-bound runtime truth public instead of
implicit.
---
## Payment
- **MCP tool calls:** [x402](https://x402.org) over USDC on Base L2
- **Pilot engagements:** the live public buy path is exact-amount USDC on Base with validated checkout capture; bank transfer, Wise, and manual exceptions remain secondary rails, and card checkout is not live
- **Support / sponsorship:** use the support page for non-customer backing of the build, infra, and open kernel
- **Chain:** Base L2 (Chain ID 8453)
- **Token:** USDC (`0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913`)
- **Wallet:** `0x82009D0fa435d490A12e0cBfBE47bf3358e47761`
---
## Managed Agent Team
Every agent is a registered entity with identity, scopes, budget, and reputation.
| Agent | Role | Purpose |
|-------|------|---------|
| **Leviathann** | Manager | Orchestrates pipeline, routes work, closes loops |
| **Atlas** | Analyst | Research, sourced findings, competitive analysis |
| **Quill** | Writer | Structured briefs, release-ready deliverables |
| **Aegis** | QA Gate | PASS/FAIL acceptance with evidence |
| **Sentinel** | Verifier | Contradiction detection, risk review |
| **Forge** | Executor | Implementation, operational steps |
| **Pulse** | Compressor | Context compression, triage |
Agents earn REP (reputation) and AUTH (authority) from accepted output. Sanctions apply for failures, fake progress, or wasted resources.
---
## Reference CI Pipeline
The CI workflow is structured as a nightly sequence. The logic is real, but public customer delivery is not being marketed as always-on hosted automation while treasury policy remains blocked.
1. **Research** — Fetch tracked sources. Watchlist competitors get priority.
2. **Extract** — Sourced findings with relevance scoring and deduplication
3. **Write** — Cited intelligence alert in structured format
4. **QA** — Multi-agent verification: source freshness, citation accuracy, quality bar
5. **Deliver** — Approved alert through the current honest path: founder-led pilot now, automated subscriber delivery when treasury policy clears
6. **Score** — Economy auto-scores agents (REP/AUTH deltas). Registry syncs.
7. **Audit** — Every step logged. Usage metered.
---
## Economy
Constitutional 3-ledger internal economy:
- **REP (Reputation)** — earned from accepted output, non-transferable
- **AUTH (Authority)** — temporary right to lead work, decays without output
- **CASH (Treasury)** — real money only (owner capital, support, customer revenue)
---
## Tech Stack
- **Runtime:** [Meridian Loom](https://github.com/mapleleaflatte03/meridian-loom) — primary execution surface for agent execution, capability dispatch, sessions
- **Platform:** Python 3.10, JSON state files, JSONL audit/metering logs
- **Proxy:** Caddy (auto-TLS)
- **Payments:** [x402](https://x402.org) + USDC on Base L2
- **Infrastructure:** VPS (Vultr), Docker sandboxing, systemd
---
*Meridian platform for governed digital labor. Built on the Meridian Constitutional Kernel and the Meridian Loom runtime. Running since 2026-03-15.*