Open Agentic Schema Framework

# Open Agentic Schema Framework

The [Open Agentic Schema Framework (OASF)](https://schema.oasf.outshift.com/) is a standardized schema system for
defining and managing AI agent capabilities, interactions, and metadata.
It provides a structured way to describe agent attributes, capabilities, and relationships using attribute-based
taxonomies.
The framework includes development tools, schema validation, and hot-reload capabilities for rapid schema development,
all managed through a Taskfile-based workflow and containerized development environment.
OASF serves as the foundation for interoperable AI agent systems, enabling consistent definition and discovery of agent
capabilities across distributed systems.

OASF is highly inspired from [OCSF (Open Cybersecurity Schema Framework)](https://ocsf.io/) in terms of data modeling
philosophy but also in terms of implementation.
The server is a derivative work of OCSF schema server and the schema update workflows reproduce those developed by OCSF.

## Features

OASF defines a set of standards for agentic AI content representation that aims to:

- Define common data structure to facilitate content standardization, validation, and interoperability.
- Ensure unique agent identification to address content discovery and consumption.
- Provide extension capabilities to enable third-party features.

## Key Concepts

At the core of OASF is the [record object](./schema/objects/record.json), which serves as the primary data structure for
representing collections of information and metadata relevant to agentic AI applications.

OASF records can be annotated with **skills** and **domains** to enable effective announcement and discovery across
agentic systems.
Additionally, **modules** provide a flexible mechanism to extend records with additional information in a modular and
composable way, supporting a wide range of agentic use cases.

## Schema Expansion and Contributions

The Open Agentic Schema Framework (OASF) is designed with extensibility in mind and is expected to evolve to capture new
use cases and capabilities.
A key area of anticipated expansion includes the definition and management of **Skills**, **Domains** and **Modules**
for AI agentic records.

We welcome contributions from the community to help shape the future of OASF.
For detailed guidelines on how to contribute, including information on proposing new features, reporting bugs, and
submitting code, please refer to our [contributing guide](CONTRIBUTING.md).

OASF can be extended with private schema extensions, allowing you to leverage all features of the framework, such as
validation.
See the relevant section in the [contributing guide](./CONTRIBUTING.md#oasf-extensions) for instructions on adding an
extension to the schema.
An OASF instance with schema extensions can be hosted, allowing you to use your own schema server for record validation.

Alternatively, records can be extended by adding arbitrary JSON objects to the `modules` list, using module names that
do not conflict with existing OASF modules.
However, this approach is the least recommended, as validation will be skipped for these modules if the record is
validated against the standard OASF schema.

### Schema Versioning and Immutability

Once a schema version is released, no changes to that version of the schema are expected, except for non-breaking fixes
such as documentation updates or minor bug corrections.
Any deletions, additions, or structural changes made after a schema version is released will be part of the next schema
version.
This immutability policy ensures backward compatibility and stability for systems that depend on specific schema
versions, while allowing the framework to evolve and improve over time through new version releases.

## Useful Links

A convenient way to browse and use the OASF schema is through the
[Open Agentic Schema Framework Server](https://schema.oasf.outshift.com) hosted by Outshift by Cisco.

To deploy the server either locally or as a hosted service, see the [server's guide](oasf-server.md) for more
information.

See [Creating an Agent Record](https://docs.agntcy.org/how-to-guides/agent-record-guide/) for more information on the
Agent Record.

The current skill set taxonomy is described in
[Taxonomy of AI Agent Skills](https://schema.oasf.outshift.com/skill_categories).

## Creating Valid OASF Records with MCP Server

The [Directory MCP Server](https://github.com/agntcy/dir/tree/main/mcp) provides powerful capabilities to help create
valid OASF agent records when configured with an LLM (such as in Cursor or other MCP-compatible IDEs).
The MCP server exposes tools and prompts that enable LLMs to generate, validate, and refine OASF records with
schema-aware assistance.

### Key Capabilities

- **Schema Discovery**:
  Query OASF schema versions and navigate domains/skills taxonomy
- **Record Generation**:
  Automatically generate complete OASF records from your codebase
- **Validation**:
  Check records against the

... (truncated)