ConceptLast verified Jul 23, 2026Owner Osyra Technical Writing

Osyra documentation

Build governed AI systems with one control plane for identity, policy, model routing, usage, audit, and verified memory.

Osyra sits between your applications and model providers. Every request follows the same authentication and policy path. The platform records evidence that operators and independent verifiers can inspect later.

Start with a working request

The fastest path is the Getting Started guide. It sends one real OpenAI-compatible request through Osyra Edge and captures the server-issued request ID used for audit and support.

Before you begin, have these values ready:

an active Osyra Edge URL;
an IAM-issued credential; and
a model ID authorized for your workspace.

Choose your path

Call models through Osyra

Add verified memory

Connect an agent or MCP client

Core concepts

Identity and policy: Osyra authenticates the caller, resolves the authoritative workspace, and applies policy before a protected operation reaches a model or service.
Model routing: applications use one governed interface while the broker selects from configured providers and models.
Evidence: request, audit, usage, and attestation records share identifiers that let operators follow one operation across services.
Verified memory: OME records agent memory in a content-addressed, tamper-evident form that can be verified independently.

SDK and API reference

Use task guides to learn a workflow. Use generated reference when you need an exact type, field, method, or error.

Migrate existing memory

Osyra provides migration guides for Pinecone, Weaviate, ChromaDB, CrewAI, LangGraph, and AutoGen.

Get help

Start with the FAQ and error catalog. Include the server-issued X-Request-ID when you report a failed API request. For a browser page error, include the displayed reference ID only when the page shows one. Otherwise include the page URL and the time the error occurred.