DialectOS

DialectOS is an open-source Spanish dialect localization MCP server and CLI tool, designed to address the problem that Spanish is not a single language but 25 regional variants. It is developed by KyaniteLabs and released under the Apache-2.0 license. Version v0.3.0 is distributed via GitHub Release archives rather than the npm registry.

Purpose

DialectOS is mainly used to translate English and other languages into 25 Spanish regional variants (such as Mexican, Argentine, Colombian, Puerto Rican, etc.), while preserving Markdown structure, code comments, and localization file formats during translation. It is suitable for AI agents, documentation teams, application developers, and support teams, helping them with regional Spanish localization.

Main Capabilities

• Multi-dialect translation: Supports 25 Spanish regional variants and understands regional differences (such as es-MX, es-ES, es-AR, es-CO, etc.).
• Structure preservation: Preserves Markdown structure, code comments, and localization file formats during translation.
• Glossary enforcement: Ensures terminology consistency.
• Quality gates: Applies adversarial quality gates to catch semantic drift, including four dimensions: token integrity, glossary fidelity, structural integrity, and semantic similarity.
• MCP server: Runs as a Model Context Protocol server, enabling AI assistants (such as Claude Desktop, Cursor, Windsurf, etc.) to natively use its 17 translation tools.
• CLI tool: Provides command-line workflows for translating, detecting, and adapting content.
• Certification and reporting: Provides incremental provider certification, adversarial dialect certification, long-document certification, and can generate customer-facing certification reports.
• Security and quality: Adversarially security-hardened, including resolving 18 CVEs, SSRF protection, circuit breakers, atomic checkpoint writes, HTML injection detection, semantic drift scoring, etc.

Usage

• Installation: DialectOS v0.3.0 is distributed via GitHub Release archives. Use the released MCP/CLI archives for agent and command-line installation; clone the repository only for local development or browser demos.
• MCP integration: Add the released MCP serviceserver to Claude Desktop, Cursor, or any MCP client. For local development, run pnpm build and point the MCP client to packages/mcp/dist/index.js.
• Browser demo: Requires a source code checkout; after running, open http://127.0.0.1:8080. The demo calls the local DialectOS backend, and the backend calls the configured provider stack.
• Recommended models: For v0.3.0, the recommended default cloud model is glm-4.5-air via the Z.ai international Anthropic-compatible endpoint; a higher-confidence option is glm-5.1; for local/offline certification, qwen3.5-9b via LM Studio can be used.
• Certification commands: Use dialect:certify for incremental provider certification;dialect:certify:adversarial for adversarial dialect certification;dialect:certify:documents for long-document certification;dialect:report Generate customer reports.
• GitHub Action: Can validate Spanish translations in CI.

Use cases

• AI agents performing Spanish regional translation.
• Documentation teams localizing documentation.
• Application developers localizing application strings.
• Support teams localizing support macros.
• Scenarios requiring validation of localization file formats and terminology consistency.

Limitations

• Primarily targets Spanish regional variants, not suitable for multilingual scenarios requiring a general-purpose machine translation API.
• The browser demo requires a source code checkout.
• Verify maturity against your own needs before production use.
• Paid authentication service is optional; the beta pilot price is $500, covering up to 10,000 source words and 5 target dialects.
• Some features (such as Portuguese dialect support, real-time collaborative translation, custom provider plugins, and VS Code extension) are on the roadmap and not yet available.