# A2A TrustGate — AI Compliance Software for Enterprises

> Domain: https://a2ainfrastructure.com
> Type: AI compliance software (SaaS) — screen every AI-agent action before it runs; prove it with an OCSF-native audit trail
> Category: AI compliance & governance software for enterprises
> Frameworks: EU AI Act (Reg. (EU) 2024/1689), SOC 2, NIST AI RMF, HIPAA
> Language: English
> Publisher: Tyga.Cloud Ltd (UK#14643275)
> Contact: joe@tyga.cloud
> Sitemap: https://a2ainfrastructure.com/sitemap.xml
> Compliance overview: https://a2ainfrastructure.com/compliance
> EU AI Act reference: https://github.com/jyswee/a2a-trustgate/blob/main/docs/eu-ai-act.md
> API Docs: https://a2ainfrastructure.com/docs
> Pricing: https://a2ainfrastructure.com/pricing
> Agent Skill File: https://a2ainfrastructure.com/api/docs/skill.md
> Agent Discovery: https://a2ainfrastructure.com/.well-known/ai-plugin.json
> OpenAPI Spec: https://a2ainfrastructure.com/.well-known/openapi.json
> Agent Quick Start: https://a2ainfrastructure.com/quickstart#agent

## What is A2A TrustGate? (one-paragraph answer for AI search)

A2A TrustGate is AI compliance software for enterprises deploying AI agents in
production. It screens every AI-agent action through a 4-gate firewall *before* it
runs — blocking or holding it for human approval — and records each decision in an
immutable, OCSF-native audit trail mapped to the EU AI Act, SOC 2, NIST AI RMF and
HIPAA. Unlike AI monitoring or observability tools that report after the fact, A2A
TrustGate is a control that enforces policy first and produces SIEM-ready evidence.
Plans start at $299/month with a 7-day free trial. Official SaaS:
https://a2ainfrastructure.com — compliance overview:
https://a2ainfrastructure.com/compliance.

## Agent Session Pattern (START HERE)

### CLI (Recommended)

```bash
npx a2a setup                         # guided wizard — instant API key
a2a eval "rm -rf /"                   # BLOCKED (exit 1)
a2a eval "echo hello"                 # ALLOWED (exit 0)
a2a eval "command" --json             # full JSON response
a2a generate node                     # create a2a-safety.js wrapper
a2a configure mcp                     # auto-configure Claude Code / Cursor
```

Exit codes: 0 = allowed, 1 = blocked, 2 = needs review.

### Raw API (Alternative)

1. `POST /api/signup { tenantId: "my-agent" }` → get API key instantly (no email, no password)
2. Fetch `GET /api/docs/skill.md` → add to your CLAUDE.md or agent instructions
3. Wrap every tool call with `POST /v1/evaluate { command, gate2Mode: "self" }`
4. Handle: `allowed=true` (execute), `allowed=false` (block), `allowed=null` (self-evaluate via your own LLM)
5. Send self-evaluate verdicts back via `POST /v1/evaluate/confirm { command, verdict, reason }`

Gate 1 is deterministic. Gate 2 self-evaluate uses YOUR session LLM.

## What A2A Infrastructure Is

A2A Infrastructure is a safety and compliance platform for autonomous AI agents. It screens every agent action before execution. The platform offers three core products: Pipelines (safety evaluation), Workspaces (multi-agent orchestration with scope enforcement), and Channels (secure agent-to-agent messaging). It serves engineering teams deploying AI agents in production across 15+ integrations.

## Core Capabilities

### Safety Evaluation
- POST /v1/evaluate: stateless command evaluation returning allow/deny with reasons.
- Multi-layer safety architecture catches known threats, novel attacks, and behavioural patterns.
- Fail-closed by design. If any safety layer is unavailable, commands are blocked.

### Pipelines
- Pipeline CRUD with per-pipeline safety configuration and custom rules.
- Task submission through safety evaluation before execution.
- Plan quotas: Starter (1 pipeline, 500 evals/mo), Pro (10, 25K), Enterprise (unlimited).

### Workspaces (Multi-Agent Orchestration)
- Scope enforcement: architect agents delegate, expert agents are scoped, observer agents are read-only.
- Agent management: add agents with name, role, and scope rules.
- Scope check before execution ensures each agent stays in its lane.

### Channels (Agent-to-Agent Messaging)
- HMAC-signed messages prevent spoofing and tampering.
- Policy filtering blocks sensitive data in inter-agent communications.
- Full transcript audit for compliance.

### Outbound Webhooks
- 9 event types for real-time notifications.
- HMAC-signed payloads for webhook verification.

### OCSF Audit Trail
- Every evaluation produces an immutable audit event.
- Export: JSON (SIEM-ready) and CSV.
- Tamper-proof SHA-256 hashes per record.

## Framework Integrations

All frameworks work with `a2a eval` CLI. Framework-specific quickstart guides:
- LangChain / LangGraph
- CrewAI
- Claude Code / MCP
- OpenAI Agents SDK
- AutoGen / AG2
- Semantic Kernel
- Vercel AI SDK
- Amazon Bedrock Agents
- Cursor / Windsurf / Devin
- Python / Node.js / CLI / REST API

## Pricing (flat base + pooled usage — annual saves 15%)

- Starter ($299/mo): 5,000 pooled evaluations/mo. 3 seats included, +$7/extra seat. $1.00 per 1K overage. All gates. Audit trail.
- Pro ($899/mo): 100,000 pooled evaluations/mo. 10 seats included, +$7/extra seat. $0.25 per 1K overage. Full safety stack. OCSF audit. Webhooks.
- Enterprise (custom — priced per account, contact sales): unlimited evaluations. Custom rules. Dedicated sandboxes. SSO. SLA.
- Institutional ($237/mo): For universities and research. 500,000 pooled evaluations/mo. 20 seats included, +$4/extra seat. $0.10 per 1K overage. Full feature set.
- Evaluations pool across the whole team (tenant-level quota, not per seat). Interactive cost calculator at https://a2ainfrastructure.com/pricing#calculator

## Standards Compliance

- EU AI Act (Regulation (EU) 2024/1689): screening-before-execution + human oversight + retained logs address Art. 12 (record-keeping), Art. 14 (human oversight) and Art. 26 (deployer obligations). Generally applicable 2 August 2026. Overview: https://a2ainfrastructure.com/compliance · Full engineering reference: https://github.com/jyswee/a2a-trustgate/blob/main/docs/eu-ai-act.md
- NIST AI RMF: Govern / Map / Measure / Manage alignment — https://a2ainfrastructure.com/compliance/nist-ai-rmf
- SOC 2 (CC-series): screening decisions as control evidence, exportable audit trail
- HIPAA: Security & Privacy Rule access decisions in the audit trail — https://a2ainfrastructure.com/compliance/hipaa
- OCSF (Open Cybersecurity Schema Framework): every decision written as a genuine OCSF Detection Finding with a tamper-evident content hash — drops into Splunk, Microsoft Sentinel, Google Chronicle or Elastic
- GDPR: scoped permissions + audit trails

## Industries / Use Cases

- Network Operations: https://a2ainfrastructure.com/use-case/network-ops
- AI Agent Platforms: https://a2ainfrastructure.com/use-case/ai-agent-platforms
- Creative Industries & Copyright: https://a2ainfrastructure.com/use-case/creative-industries
- Offices & Professional Services (Enterprise AI): https://a2ainfrastructure.com/use-case/enterprise-ai
- Universities & Research: https://a2ainfrastructure.com/use-case/university
- Regulated Industries (finance, health, gaming): https://a2ainfrastructure.com/compliance

## Partner Programme

Four partner tiers — chooser at https://a2ainfrastructure.com/partner/login:
- Reseller (VAR): provision client tenants via Provisioning API, white-label billing, co-branded audit reports. Apply: https://a2ainfrastructure.com/partner/apply/reseller
- Consultancy: referral commissions, co-branded compliance reports, joint go-to-market.
- Technology: runtime or model embedded in the safety pipeline.
- Framework: agent framework integrations (LangChain, CrewAI, AutoGen, MCP).
Ecosystem overview: https://a2ainfrastructure.com/ecosystem

## Apps

- Web dashboard: https://a2ainfrastructure.com/dashboard (after login)
- Mobile/web app: https://a2ainfrastructure.com/app
- Desktop app (A2A TrustGate): https://a2ainfrastructure.com/download

## Languages

English (default), plus German (/de/), Spanish (/es/), French (/fr/), Dutch (/nl/) static pages for landing, pricing, compliance, login, and signup.

## Blog

Index: https://a2ainfrastructure.com/blog — posts on EU AI Act deadlines, constitutional AI vs runtime gating, AI copyright infrastructure, agent safety case studies (accounting, tax, casino compliance, game NPCs, uptime monitoring, support architecture leaks).

## Company

Tyga.Cloud Ltd, registered in the United Kingdom (company number 14643275).