estonian-mcp, from Silly Geese, is an MCP server that improves Estonian language handling for AI agents by attaching linguistic tools to model context. It provides spell-checking, morphological analysis, and stylistic refinement exposed to MCP clients. Key capabilities include grammar verification, legal terminology handling, and jargon detection aimed at closing gaps in Estonian grammar. The product targets developers, writers, and legal professionals who need precise Estonian text processing within AI workflows.

What tasks can you actually use it for?

The server wraps domain resources to supply AI agents with deterministic linguistic services, not just generic suggestions. It integrates the EstNLTK toolkit and exposes more than 24 read-only tools for:

• spell-checking and orthography verification
• morphological analysis and synonym retrieval
• kantseliit (bureaucratic jargon) detection and stylistic correction

How accurate are the outputs compared to doing it manually?

Accuracy is measurable: the server posts a 96.5% score on TalTech's inflection_et benchmark for morphological tasks, which supports precise inflection and orthographic checks. Legal terminology support consults the Riigi Teataja dataset, so terminology suggestions reflect official sources. Outputs remain algorithmic; high-stakes legal or regulatory text still benefits from human review despite the strong benchmark result.

Does it require technical knowledge to get useful results?

The server expects integration into MCP-compatible workflows. It runs locally via Node.js or as a Docker container and connects to MCP clients such as Claude Desktop, Cursor, VS Code, and Zed. Local execution yields zero-latency processing and avoids external API calls, so deployment requires developer steps but provides direct control over where text is processed and stored.

A focused tool for teams that need verifiable Estonian outputs

The server is a pragmatic choice for teams that treat Estonian linguistic accuracy and on-premises processing as project requirements. Its single-language focus makes it unsuitable for pipelines that need broad multilingual coverage. Adopt the server when Estonian correctness and local data handling are gating concerns; choose a different approach if your workflow must cover many languages or non-MCP integrations.