flutterflow_cli 0.0.42

Command-line client for FlutterFlow. Export code from FlutterFlow projects and use FlutterFlow AI to build and edit your apps declaratively from the terminal.

0.0.42

• flutterflow ai init --interface fft (experimental) sets up an fft workspace instead of an SDK workspace, for coding agents in your own terminal (Claude Code, Codex, Gemini CLI, OpenCode). The project is pulled as Dart-syntax files that the agent edits, checks and pushes with the workspace's ./fft, and the workspace gets CLAUDE.md, AGENTS.md and GEMINI.md guides. fft comes from the SDK package snapshot and is compiled once per SDK build into ~/.flutterflow/fft-cli. It needs an existing project (--project when not interactive) and a new or empty folder (default <name>-fft), never an SDK workspace's, and it registers no MCP server. --interface sdk, or no --interface, sets up an SDK workspace as before.
• flutterflow ai fetch-sdk [--env <env>] installs the latest SDK package snapshot for an environment into the shared cache, verified like flutterflow ai init verifies it, and prints where it is. The desktop agent panel uses it to run the fft CLI that ships in the snapshot.
• flutterflow ai upgrade --env <env> moves an existing workspace to another SDK channel (for example from prod to beta or an enterprise region), so its SDK can match the backend it talks to. It accepts the same environments as flutterflow ai init --env.
• flutterflow ai init passes --as-library (set the workspace up for a library project) and --with-libraries auto|all|direct|none|<ids> (which of the project's libraries to pull alongside it) through to the SDK.
• A failed flutterflow ai command (an SDK compile error or a crash) no longer reports success on Windows. The router now exits 1, after printing the original code, instead of passing on an exit code of 253 or more, which pub's Windows launcher turned into a rerun and exit 0.
• The CLI runs dart pub get again when a package the SDK needs has been deleted from the pub cache, instead of failing every compile with "Error when reading …".

0.0.41

• Existing AI workspaces automatically refresh their verified SDK on the first command after a CLI update, preserving branch settings and rolling back a failed refresh.
• Running flutterflow ai init in an existing workspace refreshes SDK-owned guidance and recipes while keeping user-authored files.

0.0.40

• The CLI now reports an installation id (~/.flutterflow/install_id) with each API call, so FlutterFlow can tell one developer's machine from an API key shared across many. It is a random identifier with nothing about your machine in it — no hostname, MAC address, serial number, or user name — it is shared with flutterflow_ai so one machine counts once, and deleting the file resets it. To turn it off, set FLUTTERFLOW_NO_TELEMETRY or the cross-tool DO_NOT_TRACK to anything other than 0 — no id is minted, read, or sent, and no file is written.

0.0.39

• Fixed arrow keys not working in the flutterflow ai interactive project picker on Windows.

0.0.38

• Running flutterflow ai with no arguments (outside a workspace, on a real terminal) now launches an interactive onboarding wizard: it prompts for your API key first (reusing any saved per-base-URL credential), then shows a searchable, arrow-key project picker with "+ Create a new project" pinned on top. Pick a project and the workspace directory name is derived from it. The picker degrades to a numbered list where raw terminal mode isn't available, and to manual project-id entry when the project list can't be fetched. Set FF_AI_NO_RICH_PICKER=1 to force the numbered fallback. The existing non-interactive flags (--project, --api-key, --yes) and non-TTY invocations bypass the wizard as before.
• flutterflow ai init now fast-fails when the target path is an existing file (not just a non-empty directory), before downloading the SDK or validating your key.

0.0.37

• flutterflow ai now supports live debugging sessions: an agent running in an AI workspace can attach to the FlutterFlow Desktop app on the same machine to inspect canvas/IDE state and drive local runs, paired securely over a per-user socket (no tokens or extra prompts).

0.0.36

• flutterflow ai init now persists the validated API key to ~/.flutterflow/credentials.json (POSIX mode 0600) keyed by base URL, so subsequent inits against the same backend don't need --api-key. Pass --no-save to opt out, and FF_API_KEY in the environment still wins and is never persisted. If a saved key is later rejected by the server, the stale entry is cleared automatically and the user is re-prompted (or, in non-interactive mode, told exactly how to recover).
• Added flutterflow ai logout [--all|--base-url <url>] to list or clear saved credentials.
• The interactive API key prompt now disables terminal echo while typing.

0.0.35

• Bug fixes and improvements
• Added auto init support for Antigravity

0.0.34

• Production Release

0.0.34-beta.2

• End-to-end Windows support for flutterflow ai.

0.0.34-beta.1

• Simplified flutterflow ai init: defaults to prod and skips the interactive environment picker. Pass --env <name> to target a non-prod environment.
• Added a remote validation fallback so flutterflow ai validate and run work on all platforms.

0.0.33

• Production Release
• Relicensed from BSD 3-Clause to Business Source License 1.1, matching the FlutterFlow AI SDK. Production use is permitted in connection with FlutterFlow products and services; competing visual application builders, low-code/no-code platforms, and code generation tools are excluded. Converts to Apache 2.0 on April 22, 2055. Contact [e-mail] for commercial licensing inquiries.

0.0.32-beta.4

• Introduce the flutterflow ai namespace: a declarative Dart tool for building and editing FlutterFlow apps from your terminal. Run flutterflow ai init to scaffold a workspace, then use commands like run, validate, inspect, resources, search, plan, and mcp to drive your project. See the README for the full command list.

0.0.32-beta.3

• Bug fixes and improvements

0.0.32-beta.2

• Bug fixes and improvements

0.0.32-beta.1

• Bug fixes and improvements

0.0.31

• Add --include-export-manifest to flutterflow export-code and forward includeExportManifest to the export API so the downloaded archive can include .flutterflow/export_manifest.json.

0.0.30

• Occasionally suggest using --fix flag when it's not passed.

0.0.29

• Add support for download_url.

0.0.28

• Update path version.

0.0.27

• Update default endpoint.

0.0.26

• Update dependency versions.

0.0.25

• Update URL parsing to support more types of endpoints.

0.0.24

• Fix an issue on Windows.
• Add support for .flutterflowignore.

0.0.23

• Add more logging.

0.0.22

• Add --project-environment option.

0.0.21

• Add --as-debug option.

0.0.20

• Add --commit-hash option.

0.0.19

• Add exportAsDebug option to FlutterFlowApi.exportCode().

0.0.18

• Add deploy-firebase command.

0.0.17

• Add format option to FlutterFlowApi.exportCode().

0.0.16

• Set include_assets_map based on --[no-]include-assets when calling the API.

0.0.15

• Add environment variable for FLUTTERFLOW_ENDPOINT.

0.0.14

• Update FlutterFlowApi to throw exceptions instead of calling exit().

0.0.13

• Add --as-module option.

0.0.12

• Fix FLUTTERFLOW_PROJECT environment variable issue.
• Fix downloading assets when the parent directory does not exist.

0.0.11

• Fix --fix --no-parent-folder edge case.
• Add --environment option.

0.0.10

• Adds bug fix to duplicate archiving.

0.0.9

• Add --[no-]-parent option.
• Allow importing it inside a Dart/Flutter project.
• Allow passing project id via environment.
• Upgrade dependencies.

0.0.8

• Add --fix option.

0.0.7

• Fix handling non-existent assets.

0.0.6

• Add "branch-name" parameter to enable specifying a branch for a given project when generating code.

0.0.5

• Send API token as a header.

0.0.4

• Update API endpoint.

0.0.3

• Switch to base64.

0.0.2

• Lower min SDK version to 2.17.0.
• Allow passing SDK token as a command line argument.

0.0.1

• Initial version.