Flypic MCP tool reference

Generate an image from a text prompt. Mirrors the studio Generate tab.

Inputs

prompt (required), model, aspectRatio, quality, referenceImages[], count

Returns

Image URL once rendered, or jobId + queue position when work is queued.

Usage

Runs on your own AI provider key. Flypic does not charge per generation.

Edit an existing image using a natural-language instruction.

Inputs

imageUrl (required), prompt (required), model, quality, referenceImages[]

Returns

New image URL.

Usage

Same as generate_image.

Generate a short video from a prompt and optional starting frame.

Inputs

prompt (required unless a videoUrl drives the model), model, duration ("3"–"10"), aspectRatio, startImage, referenceImages[], videoUrl (only models with video input), isPublic

Returns

Video URL (mp4) once the model finishes — typically 30–120s.

Usage

Runs on your own AI provider key; your provider bills you directly for the video.

Higgsfield Genjutsu: keep a source video's motion and re-perform it with your character.

Inputs

videoUrl (required, 4–30s), characterImages[] (required, 1–8), prompt, resolution ("480p" | "720p" | "1080p"), isPublic

Returns

Video URL (mp4) — output length follows the source video.

Usage

Runs on your own Higgsfield key; Higgsfield bills per input second.

Paginated history of your generations, newest first.

Inputs

limit (≤100), cursor, mediaType ("image" | "video")

Returns

Array of summaries + nextCursor.

Keyword search over your past prompts. Same tokenizer the in-app autopilot uses.

Inputs

query (required), limit (≤40), mediaType

Returns

Ranked array of summaries.

Fetch a single generation by id.

Inputs

id (required)

Returns

Full generation row (prompt, urls, model, dimensions).

Your subscription plan and the features it unlocks. There is no credit balance — usage is unlimited on your own AI provider keys. (The tool name is kept for existing clients.)

Inputs

—

Returns

{ usage: "unlimited", note, plan, features[], planId, renewalDate }

Generation counts for today / this week / this month, plus your current plan.

Inputs

—

Returns

{ usage: "unlimited", plan, generations: { total, today, week, month } }

List the AI models currently enabled in Flypic.

Inputs

type ("image" | "video" | "text")

Returns

Array of { modelId, name, description, type, provider, isDefault }.

Built-in template workflows plus your own visual workflows.

Inputs

—

Returns

{ templates[], visualWorkflows[] }

Run a workflow by id. Template workflows fire their webhook server-side; visual workflows return a studio URL since their graph nodes require an in-browser session.

Inputs

workflowId (required), formData

Returns

For templates: { status, response }. For visual: { studioUrl } to open in a browser.

Check whether the user has a live signed-in Flypic studio tab attached to the autopilot remote-control channel, and what the headless fallback policy is.

Inputs

—

Returns

{ connected, fallbackMode, studioUrl, hint }

Plan an autopilot run from a natural-language request — returns the ordered UI commands + narration without executing anything. Useful for previews or for clients that want to drive the UI themselves.

Inputs

message (required), currentMode, currentPrompt, knownAliases[], attachedImages[], model

Returns

{ narration, commands[] }

Plan a request and execute it. Drives the live studio tab when one is attached; otherwise transparently falls back to a non-visible headless run (Playwright browser or a server-side service replay) so the request still completes.

Inputs

message (required), currentMode, currentPrompt, knownAliases[], attachedImages[], generateTimeoutMs, stepTimeoutMs, model

Returns

{ mode, engine, narration, timeline[], error? }

Usage

Same as the underlying generate_* tool the plan ends up running.

Poll a queued generation job to completion.

Inputs

jobId (required), timeoutMs (1s–120s)

Returns

{ status, imageUrl | text, error }

Estimate roughly what your AI provider will bill you directly for a generation with a given model. Flypic itself charges nothing per generation. Also reports whether your plan can run the model.

Inputs

modelId (required), mediaType ("image" | "video"), durationSeconds (video only)

Returns

{ usage: "unlimited", estimatedProviderCostUsd, canProceed, blockReason?, model }

Fetch a single generation job by id (status, result, error).

Inputs

jobId (required)

Returns

Full job row.

List the user's recent generation jobs, newest first, optionally filtered by status.

Inputs

status ("queued" | "running" | "succeeded" | "failed" | "cancelled"), limit (≤100)

Returns

Array of job summaries.

Cancel a queued or running generation job. Already-finished jobs can't be cancelled.

Inputs

jobId (required)

Returns

{ status: "cancelled" }

Re-run a finished (failed/cancelled) job by re-enqueueing its original parameters.

Inputs

jobId (required)

Returns

{ jobId } — the new job id.

Usage

Re-runs the original generation on your own AI provider key.

List the user's saved library assets (characters, locations, props).

Inputs

type ("character" | "location" | "prop"), limit (≤200)

Returns

Array of asset rows.

Save a new asset (character, location, or prop) to the library.

Inputs

name (required), type, description, tags[], images[] (≤12 URLs), color

Returns

The created asset row.

Delete an asset owned by the authenticated user.

Inputs

id (required)

Returns

{ ok: true }

List the user's projects with their generation counts.

Inputs

limit (≤200)

Returns

Array of project rows.

Create a new project to organize generations.

Inputs

name (required), description, color

Returns

The created project row.

Delete a project. Its generations are detached, not deleted.

Inputs

id (required)

Returns

{ ok: true }

List the user's prompt-template presets, optionally including public presets.

Inputs

includePublic, limit (≤200)

Returns

Array of preset rows.

Save a reusable prompt-template preset.

Inputs

title (required), prompt (required), description, negativePrompt, category, tags[], aspectRatio, recommendedModel, isPublic

Returns

The created preset row.

Update fields of a preset owned by the authenticated user.

Inputs

id (required), plus any updatable preset field

Returns

The updated preset row.

Delete a preset owned by the authenticated user.

Inputs

id (required)

Returns

{ ok: true }

List the user's scheduled workflow runs, optionally filtered by status.

Inputs

status ("pending" | "running" | "completed" | "failed" | "cancelled"), limit (≤200)

Returns

Array of scheduled workflow rows.

Cancel a pending scheduled workflow run owned by the authenticated user.

Inputs

id (required)

Returns

{ status: "cancelled" }