Pipelex provides an open-standard framework called MTHDS for creating executable AI methods that unify business requirements, AI agent workflows, and engineering deployment. The platform includes a runtime, gateway for model orchestration, and cloud services that let teams define, validate, and run AI workflows without traditional engineering build cycles. It supports deployment across Claude Code, Codex, on-premises infrastructure, or Pipelex Cloud.
Funding
Funding not disclosed
Founders
Product
Problem
Business teams, AI agents, and software engineers typically work from disconnected artifacts—documents, prompts, and code—leading to misalignment, costly rework, and slow delivery. Traditional approaches force a tradeoff between developer-controlled but rigid code and flexible but unreliable agent outputs, making it difficult to build AI workflows that are both deterministic and adaptable.
Solution
Pipelex introduces MTHDS, an open standard for declarative, executable AI methods that serve as a single shared artifact across business, agents, and engineering. The platform provides a reference runtime that reads typed, versioned method files and executes them reliably, with validation before runtime to catch errors early. Pipelex includes a gateway that routes inference requests to multiple model providers—including Anthropic, AWS Bedrock, and Azure OpenAI—through one unified interface with usage tracking and cost reporting. The platform supports deployment as agent tools, API endpoints, or mini-apps, and offers native plugins for Temporal and Mistral Workflows, plus a cloud option for managed execution and observability.
Target Audience
Primary users are software engineering teams and AI application builders who need to ship production-grade AI workflows, along with business stakeholders who need visibility into how AI methods are defined and executed.
Features
- MTHDS open-standard language with typed schemas, structured types, and pipe operators for declarative method definition
- Validation and dry-run capabilities that check methods before runtime execution, reducing production failures
- Pipelex Gateway providing unified access to multiple AI model providers with load balancing, fallback routing, and cost tracking
- Execution graph visualization showing method runs with full telemetry and observability for debugging
- Distributed execution and pipeline orchestration for scaling workflows across infrastructure
- CLI tools, formatter, and linter (plxt) for developer workflow integration
- Claude Code Skills Plugin and SDK support for embedding methods into agent environments
- Cloud storage and cost reporting features for production deployments