Open Responses is an open-source specification and ecosystem for building multi-provider, interoperable LLM interfaces based on the OpenAI Responses API. It defines a shared schema, and tooling layer that enable a unified experience for calling language models, streaming results, and composing agentic workflows—independent of provider.
Why Open Responses
LLM APIs have largely converged on similar building blocks—messages, tool calls, streaming, and multimodal inputs—but each provider encodes them differently. Open Responses gives builders a shared, open specification (plus reference tooling) so you can describe requests and outputs once, and run them across providers with minimal translation work.
It’s designed to be:
- Multi-provider by default: one schema that can map cleanly to many model providers.
- Friendly to real-world agentic workflows: consistent streaming events, tool invocation patterns, and “items” as the atomic unit of model output and tool use.
- Extensible without fragmentation: a stable core with room for provider-specific features when they don’t generalize yet.
Community
Open Responses is an open project built for a multi-vendor ecosystem. It’s maintained in the open, with contributions from people building across clients, routers, and model providers.
Created by: [add names here: x, y, z]
Backed by builders: a community of developers who want portability, interoperability, and a shared foundation for LLM products.
Get Started
- Read the spec to understand the core concepts: items, streaming events, and tool use.
- Review the OpenAPI reference to see the full surface area and types.
- Validate your API using the our acceptance tests.
Contributing
We’re excited to evolve the spec with a community of builders. If you’re interested in shaping interoperability across LLM providers—schemas, streaming, tooling, tests, or docs—contributions are welcome.
To understand how decisions are made and how the project is run, see the governance / technical charter.