Mantyx
Runs AI agents from code with inline prompts and tools, managing the model loop, session, observability, tracing, streaming, tool dispatch, and memory.
MANTYX is an agent runtime that lets developers run AI agents from their code by passing prompts and tools inline or calling pre-configured workspace agents. It solves the problem of integrating AI agents into existing software workflows by managing the model loop, session management, observability, tracing, streaming, tool dispatch, and memory. It sells to software developers and engineering teams in companies of all sizes, from startups to enterprises, who need to embed AI agents into their applications. It is delivered as an API and SDK, with a free tier and enterprise options, and supports multiple model providers.
Key features
- Session management
- Observability
- Tracing
- Streaming
- Tool dispatch
- Memory
- SDK for TypeScript, Go, Python
- REST API
- SSE streaming
- MCP tools for Claude and Cursor
- Configure agents in UI
- Debug in chat
- Trigger from webhooks
- Workflow DAG with branches and approvals
- Human-in-the-loop approval gates
- Support for OpenAI, Anthropic, Google, xAI
- No social media activity within the last 30 days
- Blog
- Marketplace
- API
- Docs
- Software developers
- Engineering teams
- Startups