Open Standard for AI Agents
Vendor-neutral, compliance-ready, enterprise-grade
What is Open Standard Agents?
OSSA is an open standard for AI agents. The vendor-neutral specification was created by Thomas Scola, founder of Bluefly.io. It enables interoperability across frameworks, runtimes, and organizations.
Specification Standard
OSSA is NOT a framework - it's a specification that defines the contract for agent definition, deployment, and management.
Just like OpenAPI doesn't implement APIs, OSSA doesn't implement agents. It provides the standard that implementations follow.
Framework-Agnostic
Works seamlessly with any LLM framework or SDK - LangChain, Anthropic, OpenAI, CrewAI, Langflow, AutoGen, and more.
Deploy to Kubernetes, Docker, serverless, or on-premise. OSSA is infrastructure-agnostic.
Vendor-Neutral
No vendor lock-in. Write once, deploy anywhere. Move agents between teams, organizations, and infrastructures without rewriting code.
Maintained by the open source community, ensuring long-term viability and innovation.
The OpenAPI for Agents
Just as OpenAPI standardized REST APIs, OSSA standardizes AI agent definitions
OpenAPI
Industry standard for REST APIs
- → Defines endpoints & schemas
- → Enables code generation
- → Powers API documentation
- → Vendor-neutral standard
OSSA
Industry standard for AI Agents
- → Defines agents & capabilities
- → Enables agent portability
- → Powers agent orchestration
- → Vendor-neutral standard
What's New in v0.2.9
The Foundation Release brings enterprise-grade specifications for production multi-agent systems.
Enterprise Security
- Formal security model with authentication & authorization
- Secrets management & sandboxing requirements
- FedRAMP, SOC2, HIPAA compliance profiles
Multi-Agent Orchestration
- A2A Protocol for agent-to-agent communication
- Capability URI scheme with registry format
- Instance, session, and interaction IDs
Observability
- OpenTelemetry semantic conventions
- Reasoning trace export (ReAct, CoT, ToT)
- Versioned prompt template management
Developer Experience
- Conformance testing (Basic, Standard, Enterprise)
- agents.md & llms.txt integration
- Enhanced CLI validation tools
Why Does This Matter?
Leading with portability, regulatory compliance, and vendor independence
Portability: Avoid Vendor Lock-in
Switch between AI providers (OpenAI, Anthropic, Azure) without rewriting code. Define your agent once in OSSA format, deploy anywhere. No more complete rewrites when changing frameworks.
Regulatory Compliance: Built-in Frameworks
Meet SOC2, FedRAMP, HIPAA, and GDPR requirements with standardized security models, audit trails, and data boundary controls. Compliance-ready from day one.
Vendor Independence: True Interoperability
Community-driven standard, not controlled by any single company. Works with LangChain, CrewAI, AutoGen, and any framework. Your agents, your choice.
OSSA solves this for AI agents.
One standard. Any framework. True portability. Define your agent once in OSSA format, then deploy it with LangChain, CrewAI, Anthropic, OpenAI, or any other framework. Just like OpenAPI unified REST APIs, OSSA unifies AI agents.
Get Started in Minutes
Install the CLI, create your first agent, and start building with Open Standard Agents.
1. Install CLI
npm install -g @bluefly/openstandardagents2. Create Agent
osa init my-agent
cd my-agent3. Validate
osa validate my-agent.ossa.yaml4. Export
osa export --to cursor
osa export --to langchainHow OSSA Compares
OSSA is a specification standard, not a framework. See how it enables true interoperability.
| Feature | OSSA | LangChain | AutoGen | MCP | Semantic Kernel |
|---|---|---|---|---|---|
| Vendor Neutral Not controlled by any single company | |||||
| Formal Standard JSON Schema validated specification | |||||
| Multi-runtime Works across Node.js, Python, and more | |||||
| Enterprise Governance Built-in audit trails and policy controls | |||||
| Compliance Ready SOC2, FedRAMP, HIPAA, GDPR frameworks | |||||
| Open Source Community-driven development |
OSSA
LangChain / AutoGen / Semantic Kernel
Framework-specific implementations. Open source but vendor-controlled.
MCP (Model Context Protocol)
Formal standard with multi-runtime support. Focused on context, not full agent lifecycle.
Works With Your Favorite Tools
OSSA integrates seamlessly with leading AI frameworks, platforms, and tools. Build once, deploy anywhere.
OpenAI
Anthropic
LangChainLangGraph
Langflow
Vercel AIClaude Code
DrupalAll integrations include working OSSA manifest examples in our repository.
Browse ExamplesWhy Open Standard Agents?
Build agent-based systems with confidence, knowing your agents will work across frameworks, teams, and infrastructures.
Framework-Agnostic
Works with LangChain, OpenAI, Anthropic, CrewAI, LlamaIndex, AutoGen, and kAgent. No vendor lock-in.
Portable
Move agents between teams, organizations, and infrastructures without rewriting code.
Validatable
JSON Schema validation ensures correctness before deployment. Catch errors early.
Well-Documented
Comprehensive documentation, examples, and tooling. Built for developers, by developers.
Open Source
Apache 2.0 licensed. Community-driven. Transparent development process.
Fast Integration
Export to any framework format. Import existing agents. Seamless migration paths.
Secure by Design
Built-in security patterns, authentication, and compliance features.
Observable
Built-in observability, logging, and monitoring. Track agent performance and behavior.
Ready to Get Started?
Join the community and start building with Open Standard Agents today.