AxWork Documentation
Build, deploy, and integrate AI agents with the AxWork platform.
The AxWork platform lets you create, configure, deploy, and monitor AI agents. Agents run in isolated sandbox environments (E2B + Anthropic Claude Agent SDK) with full tool, shell, and filesystem access, sharing a common memory, permissions, and MCP tool configuration.
Getting Started
Quickstart
Set up your first deployment and make an API call in minutes.
Authentication
API key and session token authentication methods.
API Reference
Interactive OpenAPI documentation for all endpoints.
Base URL
https://api.axwork.aiKey Concepts
- Agent — An AI agent configuration with a system prompt, tools, skills, and MCP servers.
- Deployment — A named endpoint (with API key) that serves a specific agent version.
- Session — A conversation between a user and an agent, supporting multi-turn interactions.
- Tool Server — An MCP server definition scoped to an organization, optionally with per-user credentials.
Guides
Streaming
One-shot, event-based, and token-by-token streaming modes.
Multi-Turn Runs
Maintain conversation context across requests.
Interactive Input
Handle agent questions and tool permission requests.
File Handling
Send files to agents and retrieve results.
AI SDK Integration
Use with useChat from @ai-sdk/react.
Per-User Identity
User-scoped memory, permissions, and MCP credentials.
Memory & Knowledge
Two-tier memory architecture with hybrid search and knowledge.