Neurelo

Every table or collection in the schema gets create, read, update, and delete endpoints, and the same definition also produces join reads and writes across related entities. Both formats ship together, so a mobile client can use a REST and GraphQL API from the same source of truth. The reference documentation is generated at the same time, and calls authenticate with an API key in the X-API-KEY header.

Some questions don't fit a CRUD route. The custom query editor lets you write SQL or MQL, run it against the data source, and publish it as its own endpoint with variables for filters. An AI query assistant takes a plain-language prompt and drafts the query using your schema as context. Neurelo says the output is usually right but not always, so the draft still needs a read-through before you deploy it.

Definitions live in JSON or YAML, which means they can be reviewed like any other source file. The visual Schema Builder writes the same file, and an entity-relationship diagram shows how tables connect. Branches and commits work the way developers expect, and a project can sync definitions with a remote Git repository, so schema changes arrive through the same review process as application code.

Each environment pins a commit to a data source, which is how a team runs an untested branch against a staging copy while production stays on the last release. Runners execute the API calls, and they run on Neurelo's cloud rather than on your servers. An environment with no traffic for seven days is stopped automatically, and core settings stay locked while runners are active.

Once observability is switched on for an environment, the dashboard records every API call and reports the slowest operations, response codes, total call volume, and usage trends by hour, day, week, or month. That's enough to spot an endpoint that started timing out after a schema change without bolting on a separate monitoring stack. Metrics stay per environment, so a test run doesn't muddy production numbers.

Schema migrations generate the SQL needed to move a database from one commit to the next, and they can be applied from the console or handed off to your own pipeline. The feature only covers PostgreSQL and MySQL, ships as a preview, and has to be enabled when the project is created. A separate generator fills an environment with mock rows at three sizes: about 10, 100, or 1,000 records per table.

Client libraries for TypeScript/JavaScript, Go, and Python wrap the generated endpoints with typed methods, so app code doesn't hand-build URLs. A published MCP server exposes schema lookups and SQL execution to Claude Desktop and other MCP clients. The API Playground inside an environment sends test calls to the live endpoints and shows the raw response before you write any integration code.