<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p" -->

---
title: When One Process Becomes Too Much: Splitting a Pipeline...
description: A practical account of splitting a tightly coupled Python pipeline (document extraction feeding fraud risk scoring) into two independent MCP servers, each...
canonical: https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: When One Process Becomes Too Much: Splitting a Pipeline into MCP Services | daily.dev
og:description: A practical account of splitting a tightly coupled Python pipeline (document extraction feeding fraud risk scoring) into two independent MCP servers, each...
og:url: https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p
og:image: https://api.daily.dev/og/posts/5eGgfE11p.png
og:image:alt: When One Process Becomes Too Much: Splitting a Pipeline into MCP Services
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# When One Process Becomes Too Much: Splitting a Pipeline into MCP Services

**[Towards Data Science](https://daily.dev/sources/tds)** · 7 min read · 0 upvotes · 0 comments

## Summary

A practical account of splitting a tightly coupled Python pipeline (document extraction feeding fraud risk scoring) into two independent MCP servers, each running as its own process with its own dependencies. The author argues the real fix is a process boundary, not MCP itself, and MCP's value is a consistent discovery/invocation interface for a growing number of services. Covers writing isolated FastMCP tools, avoiding shared utility modules that recreate coupling, using Streamable HTTP over stdio for independently deployed services, and wiring services together with langchain-mcp-adapters and LangGraph's StateGraph. Ends by noting unresolved problems: mid-execution replanning, network failure semantics, and orchestrator decision complexity as more services are added.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://towardsdatascience.com/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services>

## Questions this post answers

### Should I use MCP or plain REST to split a Python pipeline into separate services?

The process boundary matters more than the protocol. Plain REST endpoints with hand-rolled clients work fine for isolation. MCP is worth it specifically when an orchestrator needs a uniform way to reach a growing number of services under different conditions; if a service only ever has one caller, a direct internal API is simpler and a protocol layer adds no value.

_daily.dev surfaces practical takes like this for engineers weighing MCP against plain APIs when splitting services._

### Should MCP servers use stdio or Streamable HTTP transport for independently deployed services?

Streamable HTTP is the right transport for independently deployed services, since stdio ties a server's lifetime to the process that spawned it, which suits local clients but not services meant to run and scale independently. Each server then runs as its own process on its own port with no shared dependencies.

_engineers architecting multi-service pipelines can track these deployment tradeoffs on daily.dev._

### How do you keep two MCP services truly isolated instead of just cosmetically separated?

Avoid importing a shared utils.py or shared validation logic between the two servers, since that quietly recreates the exact coupling the separate processes were meant to avoid. Shared logic should either be duplicated deliberately in each service or extracted into its own third service with its own tool; isolation is a discipline maintained while writing tools, not a default benefit of using MCP.

_daily.dev helps developers designing service boundaries stay on top of patterns like this._

## Similar posts on daily.dev

- [MCP is dead. Long live the CLI](https://daily.dev/posts/mcp-is-dead-long-live-the-cli-lmlswjqmz) · Hacker News · 2 upvotes · 0 comments
- [Built My Own MCP Server in 80 Lines of Python. Now Claude Knows My Sales Pipeline.](https://daily.dev/posts/built-my-own-mcp-server-in-80-lines-of-python-now-claude-knows-my-sales-pipeline--d3qjrmhzg) · Medium · 7 upvotes · 1 comments
- [MCP standardized tools. Now we need the infrastructure around them.](https://daily.dev/posts/mcp-standardized-tools-now-we-need-the-infrastructure-around-them--qv8rzplng) · Stack Overflow Blog · 0 upvotes · 0 comments
- [The Protocol That Cleaned Up Our Agent Architecture](https://daily.dev/posts/the-protocol-that-cleaned-up-our-agent-architecture-y1dso0vxj) · Towards Data Science · 2 upvotes · 0 comments

---

Tags: [#python](https://daily.dev/tags/python), [#architecture](https://daily.dev/tags/architecture), [#microservices](https://daily.dev/tags/microservices), [#mcp](https://daily.dev/tags/mcp), [#langgraph](https://daily.dev/tags/langgraph)

[View this post on daily.dev](https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"When One Process Becomes Too Much: Splitting a Pipeline into MCP Services","url":"https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p"},"datePublished":"2026-09-09T13:08:37.784Z","dateModified":"2026-11-01T18:44:40.173Z","description":"A practical account of splitting a tightly coupled Python pipeline (document extraction feeding fraud risk scoring) into two independent MCP servers, each...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b1c5a83cb0647f1dff55d6b5c0c14730?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b1c5a83cb0647f1dff55d6b5c0c14730?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Towards Data Science","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Towards Data Science","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tds","url":"https://daily.dev/sources/tds"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"python,architecture,microservices,mcp,langgraph","timeRequired":"PT7M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Towards Data Science","item":"https://daily.dev/sources/tds"},{"@type":"ListItem","position":3,"name":"When One Process Becomes Too Much: Splitting a Pipeline into MCP Services"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/when-one-process-becomes-too-much-splitting-a-pipeline-into-mcp-services-5eggfe11p#faq","mainEntity":[{"@type":"Question","name":"Should I use MCP or plain REST to split a Python pipeline into separate services?","acceptedAnswer":{"@type":"Answer","text":"The process boundary matters more than the protocol. Plain REST endpoints with hand-rolled clients work fine for isolation. MCP is worth it specifically when an orchestrator needs a uniform way to reach a growing number of services under different conditions; if a service only ever has one caller, a direct internal API is simpler and a protocol layer adds no value. daily.dev surfaces practical takes like this for engineers weighing MCP against plain APIs when splitting services."}},{"@type":"Question","name":"Should MCP servers use stdio or Streamable HTTP transport for independently deployed services?","acceptedAnswer":{"@type":"Answer","text":"Streamable HTTP is the right transport for independently deployed services, since stdio ties a server's lifetime to the process that spawned it, which suits local clients but not services meant to run and scale independently. Each server then runs as its own process on its own port with no shared dependencies. engineers architecting multi-service pipelines can track these deployment tradeoffs on daily.dev."}},{"@type":"Question","name":"How do you keep two MCP services truly isolated instead of just cosmetically separated?","acceptedAnswer":{"@type":"Answer","text":"Avoid importing a shared utils.py or shared validation logic between the two servers, since that quietly recreates the exact coupling the separate processes were meant to avoid. Shared logic should either be duplicated deliberately in each service or extracted into its own third service with its own tool; isolation is a discipline maintained while writing tools, not a default benefit of using MCP. daily.dev helps developers designing service boundaries stay on top of patterns like this."}}]}
```

