GATEWAY INTEGRATION MODES
Choose the integration boundary you can actually govern.
Governora supports a provider-independent managed execution contract and provider-native OpenAI Responses and Anthropic Messages compatibility endpoints. The chosen mode determines which lifecycle guarantees Governora can provide.
Managed execution lets Governora control planning, decisions, evidence, capabilities and routing. Compatibility endpoints preserve familiar provider request and streaming formats, while legacy and shadow results must not be represented as fully governed.
Managed execution contract
Managed applications submit organization, workspace and actor identity; task and classification; capability envelope; opaque authorization references; output and risk; and idempotency and correlation IDs. Raw provider or connector credentials are rejected.
OpenAI Responses compatibility
Use a Governora gateway key and point the OpenAI SDK base URL at Governora /v1. Request and streaming shapes remain compatible while Governora adds run, policy, provider and reuse metadata.
- POST /v1/responses.
- store is forced to false.
- Background Responses are rejected.
- Prompt content is not ordinary gateway telemetry.
Anthropic Messages compatibility
Point the Anthropic SDK base URL at Governora and use the Governora key. Provider-native Messages and SSE streaming remain intact at the client boundary.
- POST /v1/messages.
- Messages-compatible request and stream.
- Scoped encrypted provider connection.
- Policy and provider-attempt evidence.
Declare classification and task type
Clients can send data classification and a stable task type. An explicit provider connection is only needed when administrators require it or several eligible connections remain.
- x-governora-data-classification.
- x-governora-task-type.
- x-governora-provider-connection when required.
- Automatic recognition when a task hint is absent.
Understand execution modes
Legacy preserves the existing compatibility route. Shadow builds a governed plan for comparison. Governed makes decision and evidence planning mandatory before the controlled provider boundary.
Use MCP as a companion boundary
MCP can search governed knowledge and exchange decision contracts, but a third-party AI host controls its own connectors and model loop. It is not equivalent to a managed execution.
HOW IT WORKS
Four clear steps.
Each step stays connected to the same execution. This makes authority, external work and the final outcome explainable.
- 01
Choose the mode
Select managed execution, provider compatibility or advisory companion deliberately.
- 02
Establish authority
Scope the key or identity, workspace, capabilities and authorization references.
- 03
Connect the boundary
Use the execution SDK or point the provider SDK at Governora.
- 04
Inspect the proof
Verify the plan, decisions, evidence, route, usage and receipt.
QUESTIONS
Simple questions. Direct answers.
What Governora does, what it does not do, and where it fits.
Must we rewrite our OpenAI or Anthropic integration?
Not for compatibility mode: change the key and base URL and optionally add classification headers. A fully managed execution uses the provider-independent Governora contract to express actor, risk and capabilities.
What happens when authority blocks a request?
Governora records the blocked plan and does not send the request to a tool or provider.
Does the gateway persist prompt content?
The gateway request path does not persist prompt content as ordinary gateway data. Governed application features and run evidence have their own explicit storage boundaries; confirm the required retention design during evaluation.
Map one governed AI execution.
We will map the actor, decisions, capabilities, evidence, external calls and proof with you.
Plan a pilot review →