Self-host Pipelex
v0.93.1Turn expertise into AI methods that run anywhere
- Stars
- 949
- Port
- 8081 (REST API server)
- Requires
- Python
pipelex/pipelex-apiWhat is Pipelex?
Pipelex lets software teams turn repeatable information-processing tasks — invoice extraction, contract review, HR screening, competitive intelligence — into version-controlled, typed AI methods. You describe the work in natural language; the plugin for Claude Code or Codex writes a .mthds file with explicit steps, typed inputs/outputs, and conditional routing. That single file then runs in three ways: as a FastAPI-based REST API (Docker image pipelex/pipelex-api on port 8081), as an MCP connector for ChatGPT or Claude, or as a generated Next.js web app. The runtime handles model routing across OpenAI, Anthropic, Google, Mistral, Bedrock, local models (Ollama, vLLM, llama.cpp), and adds OCR, image generation, and web search via optional extras. Configuration lives in ~/.pipelex with environment-variable overrides; authentication for the API server supports API-key or JWT modes. The code is source-available under Elastic License 2.0 (no managed-service resale), with a hosted SaaS tier and a fixed-price implementation service.
Should you self-host Pipelex?
Pipelex scores 48/100 on selfhost.directory (how scoring works) and ranks #110 of 209 in AI Agents. It has 949 GitHub stars.
It shipped 10 stable releases in the past year. The latest, v0.93.1, came out on 28 October 2026. New versions land roughly every 1 days.
Setup is advanced: plan for several services and some ongoing maintenance. An official Docker image is available. It is Other-licensed and written mostly in Python.
Its closest alternative on selfhost.directory is Activepieces (81/100, 25 258 stars). It scores higher overall, so compare the two before you pick. See Pipelex vs Activepieces side by side.
Features
- AI Method Building
- Batch Processing
- Conditional Routing
- Deterministic Orchestration
- Image Generation
- MCP Support
- Multi-Provider Model Routing
- OCR Support
- REST API
- Typed Concepts
- TypeScript SDK
- VS Code extension
- Web App Scaffolding
- Web Search
Pipelex README
FAQ
Who is the Sprint for?
Heads of AI, AI Engineering leads, and COOs in regulated or sensitive environments (banking, insurance, capital markets/PE, freelance platforms) who need repeatability, handle document volume, want code control, and can't afford a 6-month build cycle.
What happens after the Sprint?
You keep the code, methods, and Docker image. Three paths: run autonomously, keep an Extended Pipelex Support contract (monthly), or move to Pipelex Cloud (hosted platform at app.pipelex.com).
Do you offer paid implementation services?
Yes — the Pipelex AI Sprint is a fixed-price, 3-week engagement to ship three automations on your infrastructure for €20,000 (excl. VAT), with contract-backed reliability threshold and a 20% fee reduction if missed.
How is Pipelex licensed?
The MTHDS standard, SDKs, VS Code extension, and UI libraries are open source (MIT). The Pipelex runtime, runner API, MCP server, and plugin are source-available under Elastic License 2.0: you can use, build on, embed, and run them on your own infrastructure, but may not offer them as a hosted/managed service to third parties.
When is Pipelex the right choice?
For repeatable, deterministic AI tasks like invoice processing, contract analysis, report generation, and compliance review where consistent results at scale are needed. Not designed for creative exploration or open-ended tasks.
Is Pipelex an agent framework?
No. Pipelex provides deterministic methods that agents call as tools via MCP or API. It makes agents reliable at repeatable business tasks but doesn't replace your agent stack.
How does Pipelex compare to n8n, Make, Workato?
GUI-based platforms lock methods in their ecosystem. MTHDS methods are plain-text files that can be versioned in Git with semantic versioning, composed across packages with cross-package references, and deployed anywhere.
How does Pipelex compare to LangGraph, Mastra, Pydantic AI?
Those are code-first frameworks with proprietary formats where humans write Python or TypeScript. Pipelex implements the MTHDS open standard: a portable, typed language with a real package system, so methods aren't locked to any runtime.
Can other runtimes implement MTHDS?
Yes. MTHDS is an open standard with a public specification. Any platform can parse .mthds files, validate typed signatures, and execute methods. Pipelex is the reference runtime.
What is the MTHDS standard?
MTHDS is an open standard for defining, packaging, and distributing AI methods with a typed language (.mthds files) and a package system (versioned dependencies, lock files, cross-package references, federated registry at mthds.sh).
Why a new format instead of Python?
Agents need a format they can reliably generate and execute — not free-form code. .mthds files are declarative, deterministic, and portable: agents write them, humans review them, Pipelex runs them. Python stays in the runtime; the method definition stays agent-readable.
More in API Management
pRESTOpen-source, self-hosted backend that turns a PostgreSQL or MySQL database into instant REST CRUD endpoints, custom SQL routes, JWT auth and a read-only MCP API.
skrunSkrun is a self-hosted agent runtime turning Agent Skills into POST /run APIs. Supports multiple LLMs, MCP, SSE, webhooks, dashboard. MIT, Node.js, SQLite/Postgres
StrapiSelf-hosted, open-source headless CMS built with JavaScript/TypeScript that generates REST and GraphQL APIs from content models, with a customizable admin panel.
supabaseSelf-hosted open-source Postgres platform with database, auth, realtime, storage, edge functions, and vector embeddings as a Firebase alternative.
TykSelf-hosted open-source API and AI gateway proxying REST, GraphQL, gRPC, TCP and MCP traffic with auth, quotas, plugins and analytics.
unkeySource-available developer platform for modern APIs: issue and verify API keys, rate limit at a global gateway, deploy apps, and watch usage from one control plane.
wundergraphOpen-source, self-hosted lifecycle manager for federated GraphQL APIs: schema registry, breaking-change checks, router, usage analytics and tracing.
aci-devACI.dev is an open-source tool-calling platform you self-host as a backend and dev portal, wiring 600+ SaaS integrations into AI agents via function calls or unified MCP.
agnoAgno is an open-source framework for building and running self-hosted AI agent platforms with REST API, MCP server, and web UI. Supports Docker and Kubernetes.
aigne-hubAIGNE Hub is a self-hosted AI gateway for 10+ LLM providers with API key management, usage tracking, and credit billing. Built as a Blocklet.
api200Self-hosted API gateway simplifying third-party integrations with auth, caching, retries, mocks, schema watching, and MCP support via one URL.
apisixOpen-source, self-hosted API and AI gateway built on NGINX and etcd, with hot-loaded plugins, dynamic routing, load balancing and rate limiting.
0 comments