Feature and Support Matrix

What Verdictan supports, and where each piece is documented. Check here before you build on a feature, especially which request families your policies actually cover.

Customer surfaces

Interface	Use it for	Keep in mind
Console	Interactive work: organizations, agents, gateways, identity, evidence, billing	Sign in through the Console. Never copy browser credentials into API clients.
verdictan CLI	Running gateways locally and scripting control-plane work	Check each command's authentication, region, and confirmation flags.
Public API v1	Programmatic control and audit	/v1/openapi.json is the authority for schemas and permissions.
OpenAI-compatible SDKs	Standard SDKs pointed at the gateway	Policy coverage depends on the request family.

Product feature coverage

For API automation, start with the API Overview and your region's /v1/openapi.json.

Task	Console guide	CLI guide
Agents	Agents	verdictan agent
Gateway runtime	Gateway Fleet	verdictan gateway
Runtime configuration	Configurations	Policy Lifecycle
IAM, members, teams, and roles	Identity and Access	IAM, Members, Teams, and Roles
SSO and SCIM	SSO and SCIM	Control manifests export the organization's authentication policy, but not connections or SCIM.
OAuth clients	OAuth Clients	None
MFA and passkeys	MFA and Passkeys	None
Events and escalations	Overview, agent and gateway monitoring, and Inbox	verdictan events and verdictan escalation
Trail evidence	Trail and Audit Evidence	verdictan trail
History	Console History	verdictan history
Usage costs and budgets	Usage Costs and Budgets	Spend, Budgets, and API Tokens
Resource tags	Resource Tags	Control manifests support agent tags. There is no generic verdictan tag command.
Context Fabric	Agent and gateway detail views	Gateway and config workflows
Secrets	Console Secrets	Manage Secrets and Credentials
Region discovery	Console locality fields	verdictan regions

Gateway request families

Family	Route	Policy chain
Chat Completions	POST /v1/chat/completions	Input and output, JSON or SSE
Responses	POST /v1/responses	Input and output, JSON or SSE
Messages	POST /v1/messages	Input, tool, output, and audit stages, on an Anthropic-shaped body
WebSocket	GET /v1/chat/completions/ws, GET /v1/responses/ws	Each complete text frame, both directions. Binary and malformed frames are rejected, and redact verdicts don't rewrite forwarded frames.
Embeddings	POST /v1/embeddings	Not run. Authentication, provider access, and budget rules apply.
Audio transcription	POST /v1/audio/transcriptions	Not run. Authentication, provider access, and budget rules apply.
Audio speech	POST /v1/audio/speech	Not run. Authentication, provider access, and budget rules apply.
Moderation	POST /v1/moderations	Not run. Authentication and provider access apply.
Models	GET /v1/models, GET /v1/models/:model_id	Not applicable: model discovery
Published-host MCP	GET /mcp, POST /mcp	MCP tool calls, within the documented limits

Don't assume two routes get the same policy coverage because their bodies look alike. Runtime Request Families has the details for each family.

Documented provider guides

Each guide lists the credentials, endpoint, and request types for that provider:

AIML API · Alibaba Qwen · Amazon SageMaker · Azure OpenAI · Cerebras · Cloudera AI Inference · Cloudflare AI Gateway · Cloudflare Workers AI · Databricks Model Serving · DeepSeek · Fireworks AI · GitHub Models · Google AI Studio · Google Vertex AI · Groq · Hugging Face Inference · Mistral AI · OpenAI · OpenRouter · Perplexity · QuiverAI · Replicate · SambaNova · Together AI · Vercel AI Gateway · Voyage AI · xAI

Test with the exact model, request type, and streaming mode your application uses. A working text reply doesn't prove audio, tools, or structured output.

SDK and framework guides

Standard OpenAI-compatible clients:

• Node.js Integration
• Python Integration
• Java and Spring Boot Integration
• .NET Integration

Frameworks and client applications, using the same request families:

• Google ADK Integration
• Instructor Structured Output
• CopilotKit Integration
• Gradio Chat Interface
• Open WebUI

Verify a claimed capability

Before you build on a feature:

1. Find it in this matrix and in its guide.
2. Confirm the operation is in /v1/openapi.json or verdictan --help.
3. Check the environment and credentials.
4. Run one success case and one deny or failure case.
5. Check the resulting state and evidence where that feature records them.

If an operation isn't in its guide or in the public discovery document, don't build production work on it.

Next steps

• API Overview: authenticate and call the control-plane API.
• CLI Command Groups: find the CLI command for a task.
• Troubleshooting: a failed request or a missing resource.