causal_funding Eternal

Leviathan (powered by causal_funding) is a Solana-native constitutional execution and decision infrastructure stack for listing review, investment screening, and pre-trade gating.

From raw on-chain noise to an auditable BLOCK / REVIEW / ALLOW decision in minutes.

Leviathan MCP for Web4 agents: machine-consumable decision intelligence for pre-trade and pre-listing controls.

Leviathan direction: constitutional execution boundaries + machine-consumable pre-trade decisions for autonomous Web4 workflows.

Judge TL;DR

• Not a score-only tool: this demo outputs decision + evidence + operating status.
• Built for Solana speed: designed for real-time listing, investment, and pre-trade workflows.
• Agent-driven reasoning layer: converts technical evidence into operator-facing judgment.
• Production-minded quality controls for stable evaluator-facing decisions.
• Internal replay quality band: rug catch ~89%, safe block ~5%, block precision ~97%.

This is not a toy rug checker. The system is designed as a decision workflow:

1. Funding and control analysis
2. Multi-surface risk evaluation
3. Evidence packaging
4. Agent-assisted judgment synthesis
5. ALLOW / REVIEW / BLOCK decision output
6. Ongoing evaluator-facing review workflow

Web4 Agent and MCP Infrastructure Direction

As Web4 agent-based execution becomes more common on Solana, risk decisions must be machine-consumable, auditable, and repeatable.

causal_funding is positioned as decision infrastructure for both operator teams and autonomous workflows:

• Pre-trade decision gate: ALLOW / REVIEW / BLOCK
• Multi-surface risk view: funding, control, permissions, and issuer context
• LP Gold verification within the same review flow
• MCP and agent-ready evidence packaging for downstream policy engines

The goal is not to produce another alert feed. The goal is to provide a consistent decision layer that can be integrated into real listing, investing, and trading workflows.

Why Leviathan Exists (Web4 Gap)

If autonomous agents are operating capital directly, three gaps become critical:

• missing machine-verifiable execution boundaries
• missing pre-trade decision objects agents can consume directly
• missing accountable, auditable decision paths for agent actions

Leviathan addresses these gaps as a practical MVP stack:

• AEP (AI Constitution): policy guardrails and accountable execution boundaries
• Leviathan MCP: deterministic ALLOW / REVIEW / BLOCK decision contract
• Attribution core: evidence-backed risk context that powers those decisions at runtime

See docs/leviathan-web4-mvp.md for scope and MVP boundaries.

flowchart LR
    A["01 Candidate Intake"] --> B["02 Leviathan Attribution Engine"]
    B --> C["03 Risk Compression Layer"]
    C --> D["04 Agent Strategy Selection"]
    D --> E["05 AEP Policy Gate"]
    E --> F["06 Execution Ticket"]
    F --> G["07 Constrained On-Chain Execution"]
    G --> H["08 Audit, Replay Protection, State Update"]
    H --> I["09 Continuous Agent Trading Loop"]

    B1["Funding provenance<br/>Control surface<br/>Issuer profile<br/>Liquidity structure"] --> B
    C1["ALLOW / REVIEW / BLOCK<br/>Evidence summary<br/>Confidence envelope"] --> C
    E1["Identity<br/>Authority<br/>Limits<br/>Bond requirement"] --> E
    G1["Transfer / Swap<br/>Devnet or production path"] --> G

Loading

Why Solana

Solana is one of the best execution environments for real-time products: fast finality, low friction, and high throughput.

That advantage also compresses risk windows. Capital can move fast, narratives can move faster, and manual diligence often arrives too late. causal_funding is built for this exact operating reality: make explainable risk decisions at Solana speed.

What This Demo Proves

The Eternal demo is designed to prove production-minded decision quality, not just dashboard output:

• Input: token mint (pool can be auto-detected)
• Output: structured risk report + evidence package
• Decision layer: ALLOW / REVIEW / BLOCK
• Review layer: decision output, evidence context, and evaluator-facing workflow

Demo Scope

This repository is the public demo layer.

• Public demo repo: product narrative, demo behavior, sample artifacts, and weekly progress.
• Demo shell code: demo_shell/ for evaluator-facing input, rendering, and demo routing.
• Judge / pilot access: deeper evaluation flow available on request.

The repository is intentionally focused on product behavior, decision outputs, and reviewer-facing materials.

Demo Shell (Open Source Layer)

demo_shell is the upload-ready public layer for Eternal.

• Input collection and mode switching (public/judge)
• API forwarding to configured analysis backend
• Executive decision rendering
• Evidence surfaces for funding, control, permissions, and issuer context
• What-if simulation for evaluator-facing scenario analysis
• Case review and watchlist/recheck workflow
• Fallback sample for presentation safety

Quick run:

cd demo_shell
cp .env.example .env
pip install -r requirements.txt
python app.py

Judge Access

• Public demo flow: available through this repository and shared demo materials.
• Judge / pilot access: provided in controlled mode upon request.
• Contact for access:
  • Email: gauss8008@gmail.com
  • Telegram: @Leviathan_Gauss

See docs/judge-access.md for the exact request format.

Quick Demo Run (Public Repo Operator View)

Run the public shell locally:

cd demo_shell
cp .env.example .env
pip install -r requirements.txt
python app.py

Then open http://127.0.0.1:7860 and submit a mint.

Optional API call through the shell:

curl -s -X POST http://127.0.0.1:7860/api/analyze \
  -H 'Content-Type: application/json' \
  -d '{"mint":"DoBAMMqcedjoWV3m7JEU1pAzZjkQqeQzbdLUA2etbonk","judge_mode":false}'

Notes:

• When the configured backend is unreachable, the shell returns a clearly marked redacted fallback sample.
• Extended walkthroughs are available in controlled judge/pilot mode.

Expected visible output:

• Risk score + confidence
• ALLOW / REVIEW / BLOCK action
• Evidence summary and review highlights
• Decision-ready evaluator context

Current Capability Highlights

• Funding and control-surface analysis for evaluator workflows
• Multi-surface risk evaluation
• Review-ready evidence packaging
• Agent-assisted decision narrative for operator review
• Decision output aligned to real review workflows
• Public console for simulation, review, and recheck
• MCP-facing decision outputs designed for agent integration paths

Week 2 Progress

Week 2 focused on evaluator-facing product depth and runtime hardening.

• Public demo shell upgraded into a Decision Simulation & Review Console
• Controller dossier layer added for stronger operator-facing wallet context
• Token permission surface expanded for better control-risk visibility
• Metadata and issuer-footprint coverage expanded for stronger diligence quality
• Agent final-stage reliability remains under active hardening on live runs

See docs/week-2-update.md for the full update.

Week 3 Progress

Week 3 formalized Leviathan as constitutional execution infrastructure for Web4 agents, while shipping runtime quality upgrades.

• expanded AEP into a framework-agnostic constitutional kernel with a clearer case lifecycle
• added multi-action workflow coverage across trade, payment, approve, and contract call
• routed devnet trade requests toward live attribution first, with explicit fail-closed fallback
• hardened the controlled devnet execution lane and reconciliation flow
• fixed cross-token reuse overlap correctness in attribution runtime
• removed stale inbound conversion fallback when reliable pricing is unavailable
• accelerated top-source convergence probing via parallel execution

See docs/week-3-update.md for the full update.

Internal Calibration Snapshot (Current)

From the current internal provisional labeled set (rounded for public sharing):

• Rug catch rate (BLOCK + REVIEW): ~89%
• Safe block rate: ~5%
• Block precision: ~97%

This is exactly the direction we want for institutional workflows: high catch quality with controlled false blocks.

Repository Focus

This repository is intentionally product-facing.

Included here:

• Product positioning
• Demo scope
• Roadmap and integration direction
• Public-facing operating narrative

Not included here:

• Full deployment configuration
• Non-public connectors and evaluator access configuration
• Internal operating materials not required for public demo review

Who This Is Built For

• Exchange listing and risk teams
• Crypto funds and due-diligence teams
• Market makers and launch platforms
• Security and monitoring operators

Partnering

We are actively open to:

• Strategic capital partners
• Pilot integration partners (exchange/fund/workflow teams)
• Data partners for enrichment and coverage expansion

Current access is controlled for judges and selected pilot counterparts.

Repository Guide

• docs/product-positioning.md
• docs/demo-scope.md
• docs/roadmap.md
• docs/judge-access.md
• docs/demo-runbook.md
• docs/submission-checklist.md
• docs/weekly-update-template.md
• docs/video-script-weekly.md
• docs/demo-shell-architecture.md
• docs/week-2-update.md
• docs/week-3-update.md
• docs/leviathan-web4-mvp.md
• examples/sample_report_redacted.json
• demo_shell/

Status

Prepared and actively iterated for Colosseum Eternal (March 2026).