agav: MCP server for AI-driven text localization and repository tasks
agav, developed by Prapaa Ai, is an open-source Model Context Protocol (MCP) server that connects language models to a developer's local environment. It automates repository-level text localization by letting AI assistants inspect code, edit i18n files, run tests, and verify localization integrity. Key elements include terminal-native interaction, file-system access with user authorization, and a TypeScript-based codebase. The tool targets software developers, localization engineers, and teams using AI-native coding assistants who need automated internationalization workflows.
What tasks can you actually use it for?
agav targets practical repository work: managing i18n files, translating UI strings, and enforcing cross-language consistency. The server implements the Model Context Protocol (MCP) so language models receive repository context and can act inside projects. Typical outcomes include batch string updates, pull-request style edits, and scripted verification runs. Supported actions commonly performed by agents include file reads, file edits, command execution, and test invocation.
How reliable are the generated localization changes?
Reliability depends on the model used and the repository context the agent receives. agav permits assistants to execute changes and run tests from the terminal, which lets the tool produce edits that can be validated automatically. Accuracy note: translations and consistency checks reflect the underlying model's outputs and the provided prompts, so teams should include review steps for high-stakes text and ambiguous UI copy.
What file and environment inputs does it require?
The tool runs in a Node.js host and requires an MCP-compatible environment such as developer integrations that support the protocol. It is written in TypeScript and expects typical code repositories containing localization files. File-system access is gated by user authorization, so an agent can read and edit real files only after explicit permission. Non-MCP hosts and environments without Node.js support are not listed as supported targets.
Does it fit into existing developer workflows without heavy friction?
agav is terminal-native and built for integration with AI-native IDEs and assistants, which suits teams that already use agent-driven tools. Its TypeScript architecture supports extension inside JavaScript ecosystems, and being open-source allows teams to inspect how file edits are performed. Adoption caveat: teams must set up MCP-compatible tooling and establish audit processes for automated commits and test runs.
A practical tool for teams prepared to accept agent-driven localization
agav is a practical choice for software teams and localization engineers who use AI-native assistants and are prepared to audit automated edits, because it is designed around repository-aware agent interaction and open-source transparency. Teams without MCP integration or without processes to review agent-made commits may not gain the intended efficiency. A recommended practice is to combine agent edits with CI tests and human review before merging.
