AI Native Lang

Enterprise AINL: Frequently Asked Questions

Last updated: March 2026

Enterprise AINL: Frequently Asked Questions

Last updated: March 2026

General Questions

Q: What is AINL?

AINL (AI Native Language) is a programming language for deterministic AI workflows. It compiles to validated graphs with immutable audit trails, making it suitable for regulated industries.

Q: How does AINL differ from LangGraph or Temporal?

  • Deterministic by default – same input always produces same execution path
  • Compile-time validation – catch graph errors before runtime
  • Token efficient – 90-95% reduction in orchestration tokens
  • Audit-ready – JSONL execution tapes for compliance

Q: Is AINL open-source?

Yes. The core language, compiler, runtime, and adapters are Apache 2.0 licensed.

We offer commercial extensions (hosted runtimes, enterprise support, compliance tooling) under separate terms.

Technical & Integration

Q: What LLM providers are supported?

Open-source adapters exist for:

  • OpenRouter (50+ models)
  • Ollama (local models)
  • Anthropic (MCP via Claude Desktop)
  • OpenAI (direct API)
  • Custom adapters can be built for internal LLMs

Q: Can AINL integrate with our existing tools?

Yes. Adapters exist for:

  • HTTP/REST APIs – any web service
  • Databases – PostgreSQL, MySQL, SQLite
  • Message queues – RabbitMQ, Kafka (via HTTP bridge)
  • Monitoring – Prometheus, Datadog, OpenTelemetry
  • Enterprise – Salesforce, SAP, ServiceNow (custom adapters)

Q: How do you handle secrets?

AINL never logs secrets. Use environment variables or external secret managers:

node api: HTTP("call-service") {
  url: "https://api.example.com"
  headers: {
    Authorization: "Bearer ${env.API_KEY}"
  }
}

Enterprise cloud supports Vault, AWS Secrets Manager, Azure Key Vault integrations.

Q: What are the resource requirements?

  • CLI/runner: ~200MB RAM, minimal CPU
  • Hosted runtime: Scales horizontally; each tenant isolated
  • Local models (Ollama): 8GB+ RAM recommended for 70B parameter models

Compliance & Security

Q: Is AINL SOC 2 compliant?

AINL's design aligns with SOC 2 Trust Services Criteria:

  • CC6.1 – Logical access controls (RBAC)
  • CC7.2 – Monitoring of system operations (execution traces)
  • CC8.1 – Integrity of data processing (deterministic graphs)

Our enterprise offering provides automated evidence bundles for auditors. The open-core version provides the tools; you're responsible for your SOC 2 audit.

Q: Can AINL handle HIPAA-protected information (PHI)?

Yes, with proper configuration:

  • Never log PHI in execution traces (scrub or hash)
  • Use encrypted transport (HTTPS, TLS 1.2+)
  • Store PHI only in compliant databases with access controls
  • Enable strict mode to prevent accidental data leakage

Enterprise support includes HIPAA Business Associate Agreement (BAA) addendum.

Q: What about GDPR?

AINL supports GDPR through:

  • Right to erasure – delete execution traces on request
  • Data minimization – explicit data flow; no hidden LLM storage
  • Controller/processor distinction – you are data controller; AINL is processor (in hosted model)

Q: How do you ensure immutable audit logs?

Execution traces (--trace-jsonl) can be:

  • Written to append-only storage (WORM, S3 Object Lock)
  • Hash-chained for tamper evidence
  • Forwarded to SIEM (Splunk, Datadog, Elastic)

Enterprise cloud includes tamper-evident audit log service out of the box.

Performance & Reliability

Q: What is the uptime SLA for hosted runtimes?

Enterprise tier SLA: 99.9% monthly uptime (excluding planned maintenance).

This applies to the hosted runtime service only. You are responsible for your graph logic.

Q: How do you handle node failures?

AINL supports:

  • Retries – automatic with backoff
  • Circuit breakers – stop calling failing services
  • Fallback nodes – alternate logic on failure
  • Alerting – Webhook notifications on node errors

Enterprise customers get PagerDuty/Opsgenie integration.

Q: What about performance at scale?

Hosted runtimes support:

  • Autoscaling – Add workers based on queue depth
  • Priority queues – Critical graphs run first
  • Rate limiting – Per-tenant quotas
  • Cold start mitigation – Keep workers warm (configurable)

Typical latencies:

  • LLM node: 1–5s (depending on model)
  • HTTP node: <100ms (local services), 200–1000ms (external APIs)
  • Internal nodes: <10ms

Commercial & Support

Q: How much does enterprise AINL cost?

Pricing is usage-based:

  • Managed Runtime: $0.15 per 1,000 graph executions + compute time
  • Enterprise Governance: $2,000/mo base fee (includes RBAC, compliance tooling)
  • Premium Support: $5,000/mo for 24/7 SLA with 2-hour response

Exact pricing depends on volume and commitment term. Contact sales for a quote.

Q: What support tiers are available?

| Tier | Response Time | Included | |------|---------------|----------| | Community | Best effort | GitHub Discussions, Discord | | Business | 8 hours | Email support, bug fixes | | Enterprise | 2 hours (24/7) | Slack channel, phone, architecture review |

Q: Do we need to purchase support to use AINL?

No. The open-core version is free and fully functional.

Enterprise support and hosted runtimes are optional paid services.

Q: What is the difference between open-core and commercial?

| Feature | Open Core (Free) | Commercial | |---------|-----------------|------------| | Language & compiler | ✅ | ✅ | | CLI & local runner | ✅ | ✅ | | Basic adapters | ✅ | ✅ | | Self-hosted deployment | ✅ | ✅ | | Hosted runtimes (SaaS) | ❌ | ✅ | | SSO / SAML | ❌ | ✅ | | RBAC & audit logs | ❌ (you build) | ✅ (out of box) | | Compliance automation | ❌ (you configure) | ✅ (generated) | | SLA-backed support | ❌ | ✅ | | 24/7 phone support | ❌ | ✅ |

Development & Workflow

Q: Can I develop locally and deploy to hosted runtime?

Yes. Write and test locally with ainl validate and ainl run. When ready, push graphs to your hosted runtime via API or CLI.

AINL's "write once, run anywhere" philosophy means your graph is portable.

Q: How do I version control AINL graphs?

AINL files are plain text. Use Git like any code:

git add monitor.ainl
git commit -m "Add error classification node"
git push origin main

We recommend branching strategies:

  • main – production graphs
  • staging – pre-production validation
  • feature/* – experimental graphs

Q: What is the release cadence?

  • Minor releases (1.3.x): Every 2 weeks (bug fixes, small features)
  • Major releases (1.x → 2.0): Every 6-12 months (breaking changes)

We follow SemVer. Breaking changes bump major version.

Migration & Adoption

Q: Can we migrate from LangGraph?

Yes. Common migration steps:

  1. Export LangGraph state machines to AINL graphs
  2. Replace ConditionalEdges with switch statements
  3. Remove LLM call nodes where AINL nodes are deterministic
  4. Validate and test against same inputs
  5. Deploy to hosted runtime or self-host

We provide a migration guide and tools.

Q: How long does enterprise onboarding take?

Typical timeline:

  • Week 1: Training, sandbox environment, first graph deployment
  • Week 2-3: Integration with CI/CD, monitoring setup
  • Week 4: Production rollout, compliance review

Enterprise support includes implementation onboarding package.

Q: Do you offer training?

Yes:

  • Public workshops: Monthly, free for community
  • Private training: 2-day on-site or virtual for enterprise customers
  • Self-paced: Tutorials in docs/learning/ and video library

Token & Community

Q: Does enterprise require holding $AINL tokens?

No. Enterprise billing is via invoice or credit card. $AINL tokens are for community features only (governance, template marketplace, contributor rewards).

Our enterprise customers should not hold tokens for business operations.

Q: What is the token utility?

$AINL token powers:

  • Governance voting (feature prioritization)
  • Access to premium templates (token-gated)
  • Contributor rewards (templates, docs, tutorials)
  • Early feature access (staking)

Token economics are separate from enterprise SaaS revenue.

Troubleshooting

Q: Validation fails but I don't understand the error

Check the error message carefully – it includes line numbers and node names.

Common fixes:

  • Typos in variable names (AINL is case-sensitive)
  • Missing required fields in node definitions
  • Type mismatches between connected nodes

Still stuck? Ask in Discussions.

Q: Runtime errors: node fails but graph continues

By default, AINL tries to continue on node failure if possible. Use strict mode to fail fast:

ainl run graph.ainl --strict

In production, always use --trace-jsonl to see what happened.

Contact

  • Enterprise Sales: enterprise@ainativelang.com
  • Support: support@ainativelang.com (enterprise customers)
  • General: GitHub Discussions or Discord
  • Security: security@ainativelang.com (encrypted PGP available)

Need more answers? Start a discussion or contact us.