Pipelex is an open-source declarative language and devtool for building composable AI workflows that both agents and humans can use. It lets developers define multi-step agent pipelines in a readable DSL, with built-in orchestration, observability, and replay so that complex tasks become reproducible and easy to debug.
Declarative DSL
- Concise, versionable workflow definitions that are easy to review alongside application code
- Human-readable syntax that both developers and AI agents can understand and modify
- Step-level state tracking for full visibility into pipeline execution progress
Orchestration & Reliability
- Built-in orchestration engine with automatic retry and replay mechanisms
- Deterministic execution ensuring workflows produce consistent results across runs
- Structured logs for debugging complex multi-step pipelines
Pluggable Connectors
- Connectors for external APIs, databases, vector stores, and custom tools
- Compatible with multiple LLM providers for flexible model selection
- Plugin architecture allowing any custom action to be exposed as a callable tool