Toku Launches Makimoto: Open-Source AI for Asia-Pacific Data Residency

Toku, a Singapore Exchange-listed customer experience platform, has announced Makimoto, an open-source conversational AI initiative specifically engineered for Asia-Pacific data residency requirements. The first release, Makimoto Kawa, will launch as a managed transcription API on July 1, 2026, with initial hosting in Singapore.

This is not another AI model wrapped in marketing gloss. The company is open-sourcing the orchestration layer—the glue code, configuration, and APIs that allow every component in the pipeline to be swapped without rewriting the surrounding system. Under the MIT licence, developers can review, contribute to, and eventually replace each part of the stack while Toku maintains production stability for early-access users.

The announcement comes via official press release from the company, with corroboration from Tech in Asia.

Data-processing regulations across Asia-Pacific are tightening and diverging. Indonesia's Personal Data Protection Law, Vietnam's Personal Data Protection Law, and recent updates to Singapore's PDPA have all raised the weight that enterprises place on where audio, transcripts, and customer conversations are processed. For teams building voice and chat automation across the region, the primary engineering question is no longer which model is most accurate: it is which deployment satisfies each regulator while scaling across markets.

Makimoto is designed to answer that question on day one. Kawa's first release runs exclusively in Singapore, keeping customer audio and transcripts within the country and supporting in-country processing for customers subject to Singapore's PDPA and regulated-sector guidelines, including those issued by the Monetary Authority of Singapore. The platform gives teams optionality for markets such as Indonesia and Vietnam, which enforce among the strictest in-country data-processing mandates on real-time communications and transcription services.

With Kawa, customer data is processed in country, not just stored there. Every stage of the pipeline runs within a single jurisdiction, so customers always know exactly where their data is going. This matters when you're handling financial services calls or healthcare consultations—scenarios where a data breach or regulatory violation isn't just an inconvenience, it's a compliance nightmare.

Makimoto Kawa is an orchestration-first framework that assembles, configures, and runs multiple open-source components as a production-grade transcription pipeline. Both APIs run the same five-stage modular pipeline: audio resampling, noise filtering and audio enhancement, voice activity detection with speaker diarisation, speech-to-text inference, and post-processing for normalisation, temporal ordering, speaker labelling, and structured output. Operators can swap or tune any individual stage to optimise for a given language, domain, or deployment environment.

The physical reality of this architecture means developers aren't locked into a black-box service. When the speech-to-text model needs updating for better Vietnamese dialect support, you swap that component without touching the noise filtering or diarisation layers. It's like replacing a carburetor without having to rebuild the entire engine (which is how most AI vendors operate, frustratingly).

At launch, Makimoto opens the orchestration layer of Kawa as MIT-licensed source code, while its internal components remain managed by Toku. Over subsequent releases, the company will progressively open additional layers of the pipeline, working towards full transparency and community modifiability of every component. This phased approach lets the community review and contribute while maintaining production stability for early-access users.

Makimoto is operated through Makimoto Technology Pte Ltd, a wholly-owned subsidiary of the Toku Group, established to preserve community trust and transparent governance. The separation maintains a clear distinction between the enterprise and community editions of Kawa. The company commits to keeping Makimoto's open-source components permissively licensed, with no plans to adopt source-available or Business Source License-style terms.

From a business perspective, this move addresses a margin problem. Toku's gross profit margin fell from 27.4% to 24.3% in FY2025, with a larger share of sales coming from lower-margin "Usage" revenue, which reached 68.8% of total sales. The Makimoto project attempts to add a higher-margin software and AI layer to a larger connectivity-led usage business with thinner margins.

Thomas Laboulle, Founder and CEO at Toku, stated that in an era where AI capabilities evolve on a weekly cycle, the biggest risk of any technology decision is creating tomorrow's legacy today. Makimoto is the response: a composable architecture where every component can be swapped as the field evolves. The company is obsessed with the orchestration layer because models and components turn over every few months, but the layer that keeps them swappable is what makes the whole stack durable.

Alongside the early-access programme, Toku is opening ten Singapore-based positions on the Makimoto team, open to recent graduates of Singapore's universities and polytechnics. The roles span engineering, developer relations, and community work. The waiting list for early access opens at makimoto.ai.

This approach could split the AI market along geopolitical lines. It could also raise demand for locally compliant solutions in Asia-Pacific, the Middle East, and Latin America, where Toku is expanding. The company is using data residency rules as an edge over global AI providers that often process data outside the markets where customers operate.

Whether this model scales beyond the initial Singapore deployment remains to be seen. The real test comes when enterprises actually deploy this in production environments with strict compliance requirements. The open-source community will have to validate the claims, and regulators will need to accept the architecture as compliant.

For now, the waiting list is open. Whether users actually pay for the enterprise edition while the community edition evolves under open governance is the real question. Time will tell if this works in practice, not just in theory.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn