Introduction
Crosswork AI is a platform for AI-driven network operations. It provides AI agents and platform services for building, operating, and integrating AI-assisted network automation workflows. The Crosswork AI APIs and a Python Agent Software Development Kit (SDK) let developers manage and extend these capabilities from external tools or applications. The APIs use HTTPS and JSON payloads and let developers manage AI agents, model providers, retrieval context, network data connectors, platform configuration, observability destinations, and assistant conversation workflows.
Use this documentation when you need to automate Crosswork AI operations, integrate Crosswork AI with another system, build agents that run on the platform, or inspect the OpenAPI schemas for request and response details.
What Crosswork AI Exposes
The API set is organized around the main services that support AI-driven network operations:
| Area | What it Provides |
|---|---|
| Agents and tasks | Register agent packages, manage runtime state, run agent tasks, retrieve task results, and export inference output. |
| AI Assistant | Manage assistant threads, messages, feedback, suggestions, onboarding status, and notifications. |
| Large Language Model (LLM) management | Configure LLM providers and models, set defaults, manage model associations, configure OAuth 2.0 provider settings, and verify provider connectivity. |
| Retrieval and context | Store, embed, search, and retrieve document context for retrieval-augmented workflows. |
| Knowledge Graph | Ingest network topology and state, manage schema extensions, inspect schema metadata, and query changelog history. |
| Data Retrieval Adapters (DRAs) | Install and manage DRA types and instances that connect Crosswork AI agents to external network systems. |
| Model Context Protocol (MCP) Gateway | Manage tools and remote MCP server registrations exposed to agents through MCP. |
| Platform operations | Manage configuration overrides, roles, upload handling, log levels, and OpenTelemetry exporter destinations. |
| Agent SDK | Build Python agents that run inside Crosswork AI with platform clients for LLMs, tools, memory, retrieval, DRA access, observability, and testing. |
| MCP SDK | Build Python MCP servers that expose tools to Crosswork AI agents through the Model Context Protocol Gateway. |
| DRA SDK | Build Python Data Retrieval Adapters that connect Crosswork AI agents to external network systems and data sources. |
Developer Value
The Crosswork AI APIs give developers a programmable interface to the same platform services used by Crosswork AI agents and assistant workflows. Instead of manually configuring agents, model routing, data connectors, or operational settings, you can automate those changes through REST calls, CLI workflows, or SDK-based agents.
The API reference is generated from OpenAPI Specification (OAS) descriptions, so
each operation includes request parameters, request body schemas, response
schemas, and status codes. The guides provide task-level examples for common
workflows such as configuring LLM providers and using cwaictl. The Agent SDK
documentation describes how to package, deploy, test, and operate Python agents
on the Crosswork AI platform.
Common Use Cases
- Configure LLM providers, register models, set the default LLM, and route specific agents or applications to specific models.
- Build and register custom agents that use Crosswork AI platform services such as LLM completions, Knowledge Graph, retrieval context, DRA data, MCP tools, and short-term memory.
- Start asynchronous agent tasks, retrieve task results, and export inference output for reporting or downstream processing.
- Ingest network topology and state into the Knowledge Graph and extend the schema for source-specific data.
- Store and search document context so agents can retrieve relevant information during a workflow.
- Install and run Data Retrieval Adapters that bridge Crosswork AI to external network systems such as Cisco NSO and Cisco Crosswork Network Controller.
- Register custom tools or remote MCP servers so agents can discover and invoke additional capabilities.
- Automate platform administration tasks such as configuration overrides, observability exporter setup, role discovery, runtime log-level changes, and file uploads.
Where to Start
- Getting Started explains base URLs, REST methods, authentication headers, and basic API testing workflows.
- Authentication describes how to obtain and use an API token.
- API Reference lists the available OpenAPI references and downloadable OpenAPI files.
- Guides provides task-oriented examples and operational guidance.
- Agent SDK (Python) covers SDK-based agent development and deployment.
- MCP SDK (Python) covers building MCP servers that expose tools through the MCP Gateway.
- DRA SDK (Python) covers building Data Retrieval Adapters that connect Crosswork AI to external network systems.