This is the current direction, shaped by what the community needs, not a promise. There are no dates: items move by intent. Two threads run through it, making the language ready for AI-assisted work, and keeping one namespace per concern while the provider behind it stays swappable.
Where the work is focused now.
An official Marreta MCP server so an AI assistant can check and correct itself against the real toolchain, plus an "Ask Professor Martim" chat here on the site.
Run the Marreta runtime and CLI natively on Windows, without WSL, so Windows developers get the same single-binary experience as on macOS and Linux.
Run queue and topic on managed cloud messaging by swapping only the provider in configuration, not your code: Azure Service Bus, AWS (SQS, SNS, and Amazon MQ for managed RabbitMQ), and Google Cloud (Pub/Sub and Cloud Tasks). One namespace per concern, the provider behind it changes.
Declare a default on a schema field, filled by the runtime and, for stored data, written as a database column default tracked by migrations.
Run the same db. code on more engines by swapping only the provider: MySQL for another production database, and SQLite for zero-setup local development and tests, with no container to run. One namespace, swap the implementation.
Ideas we are weighing, not yet committed.
First-class recurring and scheduled work, for jobs that run on a timer rather than on a request.
OAuth2 and OIDC discovery with key rotation, beyond the current API keys and HMAC JWTs.
OpenTelemetry traces and metrics for production visibility.
The roadmap follows what people build with Marreta. If something here matters to you, or something is missing, open an issue or a discussion.