Release Notes - v2026.8.0: Unified REST and GraphQL Workspace, Data-Driven Collection Runs, and the Hoppscotch MCP Server

Hoppscotch v2026.8.0 unifies REST and GraphQL into a single workspace, makes the collection runner data-driven, improves cookie handling, ships a couple more fixes, and launches the new Hoppscotch MCP Server. 🚀

Highlights

• Unified REST and GraphQL Workspace: REST and GraphQL now live in a single tabbed workspace, so you can create, switch between, and manage both from the same place instead of separate pages. GraphQL brings its tooling with it — schema explorer and search, schema docs, headers and authorization, variables, and subscriptions with their live log — and subscriptions now also support the graphql-transport-ws protocol. A single collection can mix REST requests with GraphQL queries and mutations in one run. The unified workspace is on by default, with a toggle under Experiments in Settings.

• Data-Driven Collection Runs: Attach a CSV or JSON data file to a collection run and the runner executes the collection once per row, exposing each row’s columns as iteration variables that take priority for that iteration and are never saved. You can also run a subset of requests and export the run results as JSON.

• More Predictable Cookie Handling: A new response cookie sent without a Domain attribute now stays with the exact host that set it instead of leaking to its subdomains, and when an older Hoppscotch Agent does not return structured cookie data, the app falls back to reading the Set-Cookie response header so cookies are still captured.

• More Fixes: On desktop, an instance with a failed sign-in no longer strands you without a way back — the app checks for a recorded sign-in failure before resuming and opens the instance switcher instead. The connection-timeout dropdown also has a fixed width now.

• Introducing the Hoppscotch MCP Server: The new Model Context Protocol (MCP) server lets AI hosts like Claude Code, Claude Desktop, and Cursor work with your Hoppscotch workspace — reading and writing collections, requests, and environments, executing requests and validating responses, and generating code and documentation — with network guards and profile-based tool scoping to keep control in your hands.

This release brings some of our most requested capabilities together while continuing to strengthen the platform. Your data will be backed up automatically. If needed, see Downgrading and Restoring Backups.

👉 Read the full release blog

👉 Read the Hoppscotch MCP Server announcement