Today, we’re announcing Opik’s integration with Oracle’s Open Agent Specification, a partnership that fundamentally changes how AI teams build, test, and ship production agents.
When you build a production AI agent or workflow, the framework you pick early shapes everything that comes after it, from prompts, to logic flows, to the surrounding infrastructure. Switching frameworks means rewriting all of that. And even once you do, comparison is hard: metadata, output formats, and tool schemas differ across runtimes, so iteration slows down and teams often end up locked into their first choice. For larger teams running multiple agents, the same problem compounds at organizational scale.
That’s what this integration is built to fix.
Together, Opik and Oracle’s Open Agent Specification let you define an agent once and run, trace, and evaluate it across frameworks. The framework underneath becomes a swappable layer rather than a foundational commitment.
What is Agent Spec?
Open Agent Specification (Agent Spec) is an open source, framework-agnostic configuration language for defining AI agents. Developed by Oracle, it makes agents portable across frameworks like LangGraph, AutoGen, and WayFlow. Agent Spec captures the components that make up an agent (its LLM, prompts, tools, and flow structure) in a format that can be executed across a range of agentic runtimes. You define an agent once and run it on whichever framework fits, without rewriting prompts, tools, or orchestration logic when you switch.
Why it matters/what it unlocks
Faster experimentation and iteration. Switching frameworks no longer requires re-instrumenting your agent or rebuilding your evaluation setup. Opik traces Agent Spec-defined agents consistently across runtimes, so you can try a new framework and start comparing performance, cost, and behavior immediately.
Standardized visibility across frameworks. Agent Spec defines a consistent structure for how agent and flow executions emit traces: LLM calls, tool calls, and intermediate steps. Opik captures that structure directly, so traces look the same regardless of which runtime produced them, making it easier to debug issues and compare behavior across frameworks.
Repeatable evaluation. Because Agent Spec agents come with built-in metadata, output constraints, and a consistent structure, you can run the same Opik evaluations and test suites across different runtimes without writing custom evaluation code for each one. The same metrics apply no matter which framework the agent is running on.
A modular, framework-agnostic stack. Agent Spec and Opik are both portable and framework-agnostic, which means the agent definition layer and the observability/evaluation layer can each evolve independently. For larger teams, that makes it possible to build shared workflows and CI/CD processes that work across every framework in use, instead of maintaining a separate setup for each.
Together, Opik and Agent Spec give AI builders the flexibility to evolve their agents and their stack independently. As the framework landscape keeps expanding, teams that can move between runtimes without rebuilding everything will have a real advantage. This is how you build for that.
Getting Started
You can start using Agent Spec agent with Opik in three steps:
- Define your agent using Oracle’s PyAgentSpec SDK.
- Load it onto a runtime using the appropriate adapter for your framework of choice
- Wrap execution with Opik’s AgentSpecInstrumentor to capture traces in Opik.
The integration is available in both Opik Cloud and the open source version of Opik.
