Why We Built Makimoto as a Pipeline Instead of a Platform

When One Component No Longer Fits

Conversational AI systems change over time.

A team may begin with one transcription service and later need another for a different job.

A deployment may move into a different environment.

A new language requirement may appear.

An internal service may need to be replaced without changing the application built around it.

In a tightly coupled system, a small change can unwittingly become a full platform migration.

We wanted to avoid that.

The Design Question

As we looked at the operating problems around speech workflows, one question kept returning:

How much of the stack should be replaceable?

Our answer was not “every layer in the stack must be custom”. That would simply move complexity onto the developer instead. It was that the boundaries between layers should be clear enough to replace what matters, while keeping a coherent path through the whole workflow.

So, What Is Makimoto?

Makimoto is an open-source conversational AI initiative focused on modular building blocks for speech-driven workflows.

Instead of asking developers to adopt one fixed platform, Makimoto lets teams assemble, adapt and operate only the components they need. Some teams may begin with the speech layer. Others may use APIs, SDKs or workflow integrations. Individual parts can evolve as requirements change, without affecting other components.

A Pipeline, Not a Black Box

That decision led to Makimoto.

Makimoto treats conversational AI as a pipeline of stages rather than a single indivisible platform. Each stage has a job. The connections between stages matter as much as the components themselves.

However, this approach introduces its own responsibilities. Modular systems require clear interfaces, useful documentation, stable configuration and observable failure behaviour. Flexibility without these foundations turns into fragmentation.

What Open-Source Enables

As a developer, I would want to inspect behaviour, understand what the system depends on, change a component, run it easily, and make changes.

The fix to that user story is making Makimoto open-source.

Why APAC Still Matters

The wider problem is universal: production systems need to evolve without becoming impossible to understand.

From Architecture to Tools

Architecture only becomes useful when developers can evaluate it.

Tomorrow, we will follow one path through the Makimoto ecosystem: from Kawa, the speech layer, to the Python SDK designed to reduce the distance between discovery and a working result.

Explore Makimoto

Use these links as the next step for readers who want to explore the project or join the community.

• Join the Makimoto Discord community
• Explore Kawa on GitHub
• Explore the Makimoto Python SDK on GitHub