Takara launches Miru to cut enterprise AI coding costs

Takara on Aug. 11 launched Miru, an AI infrastructure layer designed to reduce token waste in enterprise code generation and help companies manage rising AI spend. The company says Miru can cut token waste by up to 50% and lower total AI coding costs by about 40% while keeping code quality intact.

Why it matters: - Enterprise AI coding tools are driving up invoices even as per-token prices fall. - Takara is targeting the biggest cost driver in coding agents: repository search and context loading before code is generated. - The company says Miru can help enterprises reduce AI coding spend while preserving developer productivity and governance.

What happened: - Takara announced Miru, an AI infrastructure layer built for enterprise AI code generation. - The launch took place on Aug. 11, 2026, in London. - Miru is available now through Takara’s public API and private enterprise licensing. - Takara said the product is designed to fit into existing tools and platforms without disruption.

The details: - Takara says traditional retrieval pipelines can consume 80% to 90% of an AI coding agent’s token budget before the agent writes code. - Miru moves context discovery and repository indexing from GPU-based inference to CPU infrastructure. - Takara says that shift cuts token waste by up to 50% and reduces total AI coding spend by about 40%. - The company also says Miru can save up to 60% of tokens. - Miru processes code 50 times faster and indexes 200 times faster than other options on the market, according to Takara. - The product uses an embedding model called ds1-code, which runs on ordinary CPU instances. - Takara says Miru supports sovereign deployment in local data centers or cloud environments. - The company says source code stays inside the customer’s secure infrastructure.

Between the lines: - The launch fits a broader push for standardized tokenomics led by the Linux Foundation. - That effort is focused on governance and the cost of token consumption. - The release also reflects a growing enterprise concern: AI tools such as Claude Code, Cursor and GitHub Copilot can boost productivity while also increasing token use and budget overruns. - Takara is positioning Miru as infrastructure that handles low-value context work so large language models can focus on reasoning and code generation.

What's next: - Enterprises building autonomous agent workflows across finance, operations and engineering will likely face more pressure to control token use. - Takara is betting Miru can become a building block for scaling multiple AI agents in the same workspace with clearer financial and operational returns. - The company is pushing immediate adoption through both API access and enterprise licenses.

The bottom line: - Takara is trying to turn AI coding from a token-heavy expense into a more predictable infrastructure cost.