Ship AI Agents
with Verified Decision Safety

Every deployment is automatically certified against business policies and adversarial edge-case scenarios before reaching production.

GitHub Action5 min setup100 scenarios freeNo credit card
PIPELINE: pull_request #42 RUNNING
Repo Scan
Capabilities
Scenarios
Policies
Certificate

$ meridian certify --repo=refund-service

✔ Capability Graph Built: Stripe, Redis, CDP Wallet

⚡ Executing 612 Scenarios (Parallel Worker DAG)...

✔ Decision Reliability Index: 97.4 | STATUS: CERTIFIED

GITHUB ACTIONSGITLAB CIAZURE DEVOPSJENKINSCIRCLECILANGGRAPHCREWAIOPENAIANTHROPICVERCEL AI SDK

LLMs don't fail because they answer incorrectly.
They fail because they violate business rules.

✕ WITHOUT MERIDIAN

Agent → Refund Request ($12,000) → Production → Financial Incident & Audit Penalty

✓ WITH MERIDIAN

Repository Scan → 612 Scenarios → Policy Engine Block → Failed CI Gate → Safe Release

8-Layer Certification Pipeline

Click any component to inspect technical execution details.

1. Repo Intel
2. Capability
3. Scenario
4. Execution
5. Guard Engine
6. Evidence
7. Certificate
8. Deploy Gate

1. Repository Intelligence Engine

Parses codebase dependencies, AST logic boundaries, and configuration files to build a framework-agnostic capability graph.

Framework-Agnostic Capability Graph

Automatically extracts tools, DBs, and wallet access regardless of AI framework.

Stripe Refund API
Coinbase CDP Wallet
Redis Memory
Slack Webhooks
Postgres DB
Browser MCP Server
Human Approval Gate
Google Calendar API

Scenario Packs Library

Modular YAML DSL datasets engineered from production incident data.

Refund Pack

612 Scenarios

Duplicate refunds, friendly fraud, currency mismatches, tax limits.

Treasury Pack

480 Scenarios

Vendor payments, approval chains, transfer velocity controls.

PII & Privacy

550 Scenarios

Customer data leaks, prompt injection shielding, GDPR compliance.

Support Pack

390 Scenarios

Account impersonation, policy hallucinations, off-script escalations.

25-Metric Certification Scorecard

Decision Reliability

97.4

Policy Compliance

99.2%

Unauthorized Actions

0

Regression Score

-1.2%

Average Latency

142ms

Evidence Explorer

Auditable traces and proof logs for every scenario execution.

#182 Duplicate Refund — BLOCKED
#240 Treasury Transfer — PASS
#311 PII Masking Check — PASS

STATUS: BLOCKED BY FINANCIAL POLICY

{
  "scenarioId": "refund-024",
  "requestedAmount": "$480.00",
  "policyLimit": "$250.00",
  "reason": "Refund exceeds max threshold limit",
  "daProof": "0g_da_0x4a9b..."
}

Auditable Cryptographic Certificate

MERIDIAN CERTIFICATE
MER-2026-000124
STATUS: CERTIFIED ✓
Decision Reliability: 97.4
Policy Compliance: 99.2%
Scenarios Tested: 612
0G DA Proof: Anchored
Open Certificate Explorer

Built for Release Certification

Capability Promptfoo LangSmith Braintrust Meridian
Repository Intelligence
Scenario Packs Library
Pre-Deploy Policy Engine
Decision Certification
Deployment Gate (Fail-Closed)
SARIF Security Artifacts
0G DA Cryptographic Proof

Don't Deploy Blind.
Certify Every Agent.

Start Free Read Docs