google-flow-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| FLOW_MCP_CDP_URL | No | The Chrome DevTools Protocol URL to attach to. | http://127.0.0.1:9222 |
| FLOW_MCP_OUTPUT_DIR | No | Default directory where downloaded images are saved. | ~/Downloads/flow-mcp |
| FLOW_MCP_AUTO_LAUNCH | No | Set to 0 to require a Chrome you started yourself. | 1 |
| FLOW_MCP_CHROME_PATH | No | Path to the Chrome executable. Auto-detected for macOS, Windows and Linux. | |
| FLOW_MCP_PROFILE_DIR | No | Directory of the dedicated Chrome profile used by the server. | ~/.flow-mcp/chrome-profile |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| flow_statusA | Connect to the Flow Chrome window (launching it on first use) and report whether you are signed in. Call this first; if signedIn is false, ask the user to sign into Google in the Chrome window that opened. |
| flow_get_optionsA | Read the image models, aspect ratios and image counts Flow currently offers this account, plus the current selection. Values are read live from the UI; use them verbatim in flow_generate_images. |
| flow_generate_imagesA | Set aspect ratio, model and number of images in Google Flow, submit the prompt, and wait until the images finish. Returns the new media ids. Pass download_size to also download them in the same call. Refuses to run when Flow says it would cost more than max_credits. |
| flow_download_imagesA | Download images from a Flow project by media id. 1K is the original file; 2K and 4K are upscaled by Flow on request and can take a few minutes. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 4 tools
Each tool maps to a distinct step in a linear pipeline (connect/sign-in, read options, generate, download), so misselection is unlikely. The one soft overlap is that flow_generate_images can also download when given download_size, duplicating part of flow_download_images' purpose.
All names share a clear flow_ prefix in snake_case, and three of four follow a verb_noun pattern (get_options, generate_images, download_images). flow_status deviates by omitting the verb, a minor inconsistency against an otherwise predictable scheme.
Four tools is a tight, purposeful set matching a narrow browser-automation workflow, with no filler tools. It sits at the low edge of the comfortable range, but each tool earns its place as a pipeline stage.
The happy path (status, options, generate, download) is fully covered with no dead ends for a straightforward generation run. Minor gaps exist around recovery and discovery: no way to list existing projects/media ids or re-check an in-flight generation, so a timed-out generate leaves the agent with no fallback.