Convex

Local deployments now upgrade to a new backend version in place, bypassing snapshot export and import, so upgrades no longer depend on data size or prompt about transferring it. The Convex MCP server now supports the stateless MCP protocol (2026-07-28 MCP specification), and a new getServiceToken function generates service tokens for first-party Convex services.

Local deployments now upgrade to a new backend version in place, skipping the snapshot export and import that previously scaled with data size. The Convex MCP server supports the stateless MCP protocol, and a new getServiceToken function generates service tokens for first-party services.

Adds a new @convex-dev/no-schema-import-cycle rule, enabled by default as an error, that detects when schema.ts imports files that themselves depend on the schema.

Development-environment tokens like npx convex dev can now be scoped to narrower permissions than the full user, preventing agents from pushing to production inadvertently.

schema.doc(tableName) now returns the validator for whole documents of a table, including id and creationTime system fields, with schema.id() and docValidator() added as companions. ConvexSiteUrl and ConvexCloudUrl are available on env by default in generated server types, and fixed a referential stability bug in ConvexProviderWithAuth plus a misleading npx convex login status report.

Public deployment APIs for continuous streaming export of data are now available.

All Convex functions can now access HTTP headers, useful for checking IPs or custom client headers like version or agent—a long-requested feature with 48 upvotes.

A new @convex-dev/no-top-of-hour-crons rule (enabled by default as a warning) flags cron jobs scheduled exactly on the hour. @convex-dev/no-collect-in-query and @convex-dev/explicit-table-ids now work without type-aware linting, and error messages recommend enabling it when disabled.

New CLI commands npx convex deployment usage and npx convex deployment usage-limits let you view resource usage and edit limits. Mutations can now reference their upcoming commit timestamp via ctx.db.vars.commitTs (with a matching v.commitTs() validator), useful for FIFO queues, and hourly cron jobs can omit minuteUTC for auto-assigned start times.

Business customers now pay a lower overhead cost for paused deployments, previously charged the same as unpaused ones, better supporting teams with many idle deployments.

The Convex dashboard now supports quickly searching and navigating between projects, deployments, and data.

You can now quickly switch between projects and deployments on the Convex dashboard using search and navigation.

New public API enables continuous streaming export of Convex data to other data locations, supporting greenfield and brownfield projects where data spans multiple stacks.

Convex Dashboard and CLI now support logging in with a Password or Passkey, in addition to Google and GitHub, along with MFA for more secure access.

A public API now enables continuous streaming export of Convex data to other data locations, ideal for grayfield projects spanning multiple stacks.

Actions can now run for substantially longer durations, helping teams handle heavy imports, large batch jobs, and slow third-party integrations more reliably.

Fixed a bug where the codegen did not sort module paths consistently with other platforms on Windows. This completes a fix partially applied in 1.42.2.

You can now configure granular usage limits for individual deployments, such as "Allow 50 GB of DB I/O per month."

Mutations and actions can now read the raw authentication token via authToken in ctx.meta.getRequestMetadata(). Fixed a circular import in convex/browser affecting ConvexHttpClient, a ConvexProviderWithClerk bug that ignored session changes, and a Windows codegen sorting inconsistency.

Convex dashboard now displays a visual diagram of a deployment's schema.