Black Forest Labs connector for AI agents

Tools your creative agent reaches for on Black Forest Labs, scoped per user.

CALL ANY TOOL

23 tools for image generation in Black Forest Labs: generate images with FLUX.2, FLUX 3, and FLUX1.1, edit with Kontext, fill, expand, outpaint, and erase regions, generate, edit, and upscale video, and poll results.

blackforestlabs_deblur_image

Deblur image

Sharpen a blurry image with FLUX Deblur while keeping the scene, identity, colors and composition. Returns the task id and polling_url (plus cost in credits when reported); the image is not in this response. Use this to fix blur without a prompt. Use blackforestlabs_generate_image for other instruction-based edits. Then call blackforestlabs_get_result with the returned id until status is Ready.

Parameters

Name

Type

Required

Description

image

string

Required

Blurry image to sharpen, as a public http(s) URL or a base64-encoded image.

output_format

string

Optional

File format of the generated image: jpeg, png, webp. Omit to use the endpoint default (png). One of: `jpeg`, `png`, `webp`.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-5 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

seed

integer

Optional

Optional seed for reproducible results. The same seed, prompt and settings give a similar image; omit for a random seed. Example: 42

blackforestlabs_edit_image_kontext

Edit image kontext

Edit an image from a text instruction, or create one from text, with FLUX.1 Kontext [pro] or [max]. Returns the task id and polling_url (plus cost in credits when reported); the image is not in this response. Use this when you specifically want Kontext, for example to keep an existing Kontext workflow. BFL recommends blackforestlabs_generate_image (FLUX.2) for new editing work. Use blackforestlabs_fill_image to repaint only a masked area. Then call blackforestlabs_get_result with the returned id until status is Ready.

Parameters

Name

Type

Required

Description

prompt

string

Required

Edit instruction for the input image, or a description of a new image when no input image is given. Example: Change the car color to matte black and keep everything else the same.

additional_images

array

Optional

Up to 3 extra reference images (URLs or base64) sent as input_image_2 to input_image_4. BFL marks multi-reference Kontext as experimental.

aspect_ratio

string

Optional

Output aspect ratio between 21:9 and 9:21, as width:height. Omit to keep the input image's ratio (or the model default for text-to-image). Example: 16:9

input_image

string

Optional

Image to edit, as a public http(s) URL or a base64-encoded image. Omit to create an image from the prompt alone.

model

string

Optional

Kontext model: flux-kontext-pro (default) or flux-kontext-max (higher quality; BFL allows at most 6 concurrent max requests). One of: `flux-kontext-pro`, `flux-kontext-max`. Default: `flux-kontext-pro`.

output_format

string

Optional

File format of the generated image: jpeg, png, webp. Omit to use the endpoint default (png). One of: `jpeg`, `png`, `webp`.

prompt_upsampling

boolean

Optional

Set to true to let BFL rewrite the prompt for more creative results. Off by default.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-6 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

seed

integer

Optional

Optional seed for reproducible results. The same seed, prompt and settings give a similar image; omit for a random seed. Example: 42

blackforestlabs_edit_video

Edit video

blackforestlabs_erase_object

Erase object

blackforestlabs_expand_image

Expand image

blackforestlabs_fill_image

Fill image

blackforestlabs_generate_image

Generate image

Generate or edit an image with FLUX.2 [pro], [max] or [pro] preview from a text prompt and up to 8 optional reference images. Returns the task id and polling_url (plus cost in credits when reported); the image is not in this response. Use this as the default for text-to-image and instruction-based edits. Use blackforestlabs_generate_image_flux3 for FLUX 3 Image, blackforestlabs_generate_image_flux2_flex for typography and step/guidance control, and blackforestlabs_generate_image_flux2_klein for the fastest, cheapest renders. Then call blackforestlabs_get_result with the returned id until status is Ready.

Parameters

Name

Type

Required

Description

prompt

string

Required

Text description of the image to generate, or the edit to apply when input_images are given. Refer to reference images by position (image 1, image 2). Hex colors such as #FF5733 are followed precisely.

disable_pup

boolean

Optional

Set to true to generate from the prompt exactly as written. FLUX.2 [pro] and [max] rewrite (upsample) the prompt by default for better results.

height

integer

Optional

Output height in pixels (minimum 64; FLUX.2 outputs up to 4 megapixels). Omit to let BFL choose: the first reference image's size when editing, otherwise the model default. Set together with width. Example: 810

input_images

array

Optional

Reference images for editing or multi-reference generation, in order (image 1, image 2, ...). Each entry is a public http(s) image URL or a base64-encoded image. Up to 8 images. Refer to them in the prompt by position, for example 'the sofa from image 2'. Omit for plain text-to-image.

model

string

Optional

FLUX.2 model to run. flux-2-pro is BFL's recommended default for generation and editing; flux-2-max gives the highest quality and strongest edit consistency; flux-2-pro-preview gets BFL's newest quality and speed updates first (prefer flux-2-pro for stable production use). One of: `flux-2-pro`, `flux-2-max`, `flux-2-pro-preview`. Default: `flux-2-pro`.

output_format

string

Optional

File format of the generated image: jpeg, png, webp. Omit to use the endpoint default (jpeg). One of: `jpeg`, `png`, `webp`.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-5 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

seed

integer

Optional

Optional seed for reproducible results. The same seed, prompt and settings give a similar image; omit for a random seed. Example: 42

width

integer

Optional

Output width in pixels (minimum 64; FLUX.2 outputs up to 4 megapixels). Omit to let BFL choose: the first reference image's size when editing, otherwise the model default. Set together with height. Example: 1440

blackforestlabs_generate_image_finetuned

Generate image finetuned

blackforestlabs_generate_image_flux11_pro

Generate image flux11 pro

blackforestlabs_generate_image_flux11_ultra

Generate image flux11 ultra

blackforestlabs_generate_image_flux1_dev

Generate image flux1 dev

blackforestlabs_generate_image_flux2_flex

Generate image flux2 flex

Generate or edit an image with FLUX.2 [flex], with control over steps and guidance and up to 8 reference images. Returns the task id and polling_url (plus cost in credits when reported); the image is not in this response. Use this for typography, text rendering and preserving small details, or when you need to tune steps and guidance. Use blackforestlabs_generate_image for general FLUX.2 [pro]/[max] work. Then call blackforestlabs_get_result with the returned id until status is Ready.

Parameters

Name

Type

Required

Description

prompt

string

Required

Text description of the image, or the edit to apply to the reference images. Quote any text to render, for example: A poster that says "OPEN LATE" in bold red letters.

guidance

number

Optional

Guidance scale, 1.5 to 10 (server default 5). Higher values follow the prompt more closely at some cost to realism.

height

integer

Optional

Output height in pixels (minimum 64). Omit for the default size. Set together with width. Example: 1024

input_images

array

Optional

output_format

string

Optional

File format of the generated image: jpeg, png, webp. Omit to use the endpoint default (jpeg). One of: `jpeg`, `png`, `webp`.

prompt_upsampling

boolean

Optional

Whether BFL rewrites the prompt before generating (server default true). Set false to use the prompt exactly as written.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-5 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

seed

integer

Optional

Optional seed for reproducible results. The same seed, prompt and settings give a similar image; omit for a random seed. Example: 42

steps

integer

Optional

Generation steps, 1 to 50 (server default 50). More steps give more detail but take longer.

width

integer

Optional

Output width in pixels (minimum 64). Omit for the default size. Set together with height. Example: 1024

blackforestlabs_generate_image_flux2_klein

Generate image flux2 klein

blackforestlabs_generate_image_flux3

Generate image flux3

blackforestlabs_generate_image_flux3_klein

Generate image flux3 klein

Generate or edit an image with FLUX 3 Image Klein, a smaller, faster FLUX 3 model, with up to 10 reference images. Returns the task id and polling_url (plus cost in credits when reported); the image is not in this response. Use this for faster, cheaper FLUX 3 output or exact pixel sizes. Use blackforestlabs_generate_image_flux3 for the full FLUX 3 Image model. Then call blackforestlabs_get_result with the returned id until status is Ready.

Parameters

Name

Type

Required

Description

prompt

string

Required

Free-form description of the image, or the edit instruction when images are given.

aspect_ratio

string

Optional

Output aspect ratio. auto (server default) keeps the first reference's aspect for edits or picks one from the prompt for text-to-image. Ignored when width and height are set. One of: `auto`, `21:9`, `2:1`, `16:9`, `3:2`, `7:5`, `4:3`, `5:4`, `1:1`, `4:5`, `3:4`, `5:7`, `2:3`, `9:16`, `1:2`, `9:21`.

height

integer

Optional

Exact output height in pixels, 64 to 6256; set together with width. Overrides resolution and aspect_ratio.

images

array

Optional

Optional reference images, each a public http(s) URL or base64 image, 1 to 10 total. The first is the image being edited and sets the default output size.

output_format

string

Optional

File format of the generated image: jpeg, png. Omit to use the endpoint default (jpeg). One of: `jpeg`, `png`.

resolution

string

Optional

Resolution tier: 768sq, 1k, 1.5k, 2k or 4k. Omit for 1k (text-to-image) or the first reference's size (edits). Ignored when width and height are set. One of: `768sq`, `1k`, `1.5k`, `2k`, `4k`.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-4 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

seed

integer

Optional

Seed for reproducibility, 0 to 4294967295; random when omitted. Example: 42

steps

integer

Optional

Denoising steps, 1 to 16 (server default 16).

width

integer

Optional

Exact output width in pixels, 64 to 6256; set together with height. Overrides resolution and aspect_ratio. width x height may be at most 16,822,272 pixels.

blackforestlabs_generate_video

Generate video

Generate a video with synchronized audio using FLUX 3, from text (t2v), keyframe images (i2v), a video to continue (v2v), or a saved draft (draft_enhance). Returns the task id and polling_url; when Ready, get_result returns signed video URLs in result.samples (and draft_caches when draft is true). Use this to create new video. Use blackforestlabs_edit_video to change an existing clip and blackforestlabs_upscale_video to raise its resolution. Then call blackforestlabs_get_result with the returned id until status is Ready; video renders take minutes.

Parameters

Name

Type

Required

Description

mode

string

Required

Generation mode. t2v: text to video. i2v: your keyframes become frames of the video (needs keyframes). v2v: continue from the end of start_video (needs start_video). draft_enhance: render a previous draft at full quality (needs draft_cache; prompt and other settings are ignored except resolution). One of: `t2v`, `i2v`, `v2v`, `draft_enhance`.

aspect_ratio

string

Optional

Output aspect ratio, or auto (server default) to choose from the prompt and references. Not used by draft_enhance. One of: `auto`, `21:9`, `2:1`, `16:9`, `4:3`, `1:1`, `3:4`, `9:16`, `9:21`.

draft

boolean

Optional

Draft mode: render a fast preview and return a draft_cache you can later pass back with mode draft_enhance for the full-quality render. Off by default. Not used by draft_enhance.

draft_cache

string

Optional

draft_enhance only. The draft bundle from a previous draft run: the base64 .bin downloaded from result.draft_caches, or its http(s) URL while that URL is still valid.

duration

integer

Optional

Video length in whole seconds: 5 to 20 for t2v and i2v, 5 to 15 for v2v. Omit to let the model fit the content (auto).

generate_audio

boolean

Optional

Generate synchronized audio with the video (server default true). Not used by draft_enhance.

keyframe_times

array

Optional

i2v only, optional. Time in seconds for each keyframe, in the same order as keyframes and in increasing order, e.g. [0, 4.5, 9]. When set, keyframes are sent as [seconds, image] pairs and duration must be an integer.

keyframes

array

Optional

i2v only. Images that become frames of the video, each a public http(s) URL or base64, 1 to 10. One image starts the video, two start and end it, more are spread evenly between first and last (3 or more need an integer duration). Images must be at least 256x256.

prompt

string

Optional

Description of the video: action, camera, lighting and any dialogue or sound. Required for t2v, i2v and v2v; ignored for draft_enhance.

resolution

string

Optional

Resolution class: hd (server default for new videos), or fhd, qhd, uhd for a higher-resolution result finished by the video upsampler. draft_enhance defaults to fhd. One of: `hd`, `fhd`, `qhd`, `uhd`.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-4 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

start_video

string

Optional

v2v only. Video to continue, as a public http(s) URL or base64 mp4; the new clip carries on from its final frames.

blackforestlabs_get_credits

Get credits

blackforestlabs_get_finetune_details

Get finetune details

blackforestlabs_get_result

Get result

blackforestlabs_list_finetunes

List finetunes

blackforestlabs_outpaint_image

Outpaint image

Place an image on a larger target canvas and generate the surrounding area with FLUX Outpainting. Returns the task id and polling_url (plus cost in credits when reported); the image is not in this response. Use this to reach a target output size or aspect ratio in one call. Use blackforestlabs_expand_image to add exact pixel counts to individual edges. Then call blackforestlabs_get_result with the returned id until status is Ready.

Parameters

Name

Type

Required

Description

height

integer

Required

Target canvas height in pixels (minimum 64). Example: 1080

input_image

string

Required

Image to extend, as a public http(s) URL or a base64-encoded image.

width

integer

Required

Target canvas width in pixels (minimum 64). Example: 1920

auto_crop

boolean

Optional

Set to true to crop the input image where it extends past the canvas; otherwise that returns an error. Off by default.

disable_pup

boolean

Optional

Set to true to skip BFL's image-aware prompt upsampler for lower latency, at some cost to quality.

mode

string

Optional

Quality/speed trade-off: high (server default) for the best fidelity, fast for quicker results that suit landscapes, backgrounds, textures and products. One of: `high`, `fast`.

output_format

string

Optional

File format of the generated image: jpeg, png, webp. Omit to use the endpoint default (png). One of: `jpeg`, `png`, `webp`.

prompt

string

Optional

Experimental optional guidance for the new area; the input image's content is the main signal. Example: a busy city street at dusk.

reference_offset_x

integer

Optional

Left offset in pixels of the input image's top-left corner on the canvas; negative values allowed. Omit to center horizontally.

reference_offset_y

integer

Optional

Top offset in pixels of the input image's top-left corner on the canvas; negative values allowed. Omit to center vertically.

safety_tolerance

integer

Optional

Moderation tolerance for inputs and outputs, where 0 is the strictest. BFL documents 0-5 for this endpoint and defaults to 2; a value above the endpoint's maximum returns a 422 error. Example: 2

blackforestlabs_upscale_video

Upscale video

blackforestlabs_virtual_try_on

Virtual try on

Build your Agent

Same auth pattern across LangChain, OpenAI, Anthropic, and Google ADK.

import { ScalekitClient } from "@scalekit-sdk/node";
import { createReactAgent } from "@langchain/langgraph/prebuilt";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);

// Black Forest Labs tools scoped to this user
const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["blackforestlabs"], toolNames: [
    "blackforestlabs_generate_image",
    "blackforestlabs_edit_image_kontext",
    "blackforestlabs_get_result"] },
  pageSize: 100,
});

const agent = createReactAgent({ llm, tools });
await agent.invoke({ messages: [{ role: "user", content: "Generate a hero image of a greenhouse at dawn and return the result" }] });

import OpenAI from "openai";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);
const openai = new OpenAI();

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["blackforestlabs"] }, pageSize: 100,
});

const res = await openai.chat.completions.create({
  model: "gpt-5",
  messages: [{ role: "user", content: "Generate a hero image of a greenhouse at dawn and return the result" }],
  tools,
});

// Execute the tool call with the user's vaulted Black Forest Labs API key
await sk.tools.executeTool(res.choices[0].message.tool_calls[0], "user_123");

import Anthropic from "@anthropic-ai/sdk";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);
const anthropic = new Anthropic();

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["blackforestlabs"] }, pageSize: 100,
});

const msg = await anthropic.messages.create({
  model: "claude-sonnet-5",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Generate a hero image of a greenhouse at dawn and return the result" }],
  tools,
});

// Tool call runs with the user's vaulted Black Forest Labs API key
await sk.tools.executeTool(msg.content, "user_123");

import { Agent } from "@google/adk/agents";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["blackforestlabs"] }, pageSize: 100,
});

const agent = new Agent({
  name: "creative_agent",
  model: "gemini-2.5-pro",
  instruction: "Black Forest Labs tools scoped to this user",
  tools,
});

await agent.run("Generate a hero image of a greenhouse at dawn and return the result");

Try these prompts

Copy any prompt into your agent. Each maps directly to a Black Forest Labs tool. Click to copy, paste into your agent, done.

Generate images

Copy the prompt

Copied

Generate a product shot of a ceramic mug on a walnut desk with FLUX 3.

Copy the prompt

Copied

Generate a 4 megapixel landscape poster with FLUX1.1 [pro] Ultra.

Copy the prompt

Copied

Generate an image of our office mascot with our brand finetune.

Edit images

Copy the prompt

Copied

Change the background of this photo to a beach at sunset.

Copy the prompt

Copied

Erase the coffee cup from this image using this mask.

Copy the prompt

Copied

Expand this square image to a wide banner by adding sky on both sides.

Video and account

Copy the prompt

Copied

Generate a short video of waves hitting a pier with FLUX 3.

Copy the prompt

Copied

Upscale this video clip 2x toward 4K.

Copy the prompt

Copied

How many credits are left on our Black Forest Labs account?

SEE HOW AUTH WORKS

Each user connects their own Black Forest Labs API key once, and Scalekit sends it with every call. Keys stay vaulted, every call is scope checked, and every action is logged.

Authorize

Your user connects

Black Forest Labs

once. We tie it to their identity and the meetings they approved — no shared bot account, no org-wide access

Who:

user ‘A’

when:

Once per user

access:

Limited to user

Store

Their

Black Forest Labs

token lives in a vault scoped to them. User A's meetings are never reachable by an agent acting for user B, even on the same connection

vault:

encrypted

scope:

per-user

tokens:

auto-refreshed

Resolve

When your agent calls a

Black Forest Labs

tool, we fetch the right token server-side. It never touches your agent, never appears in the LLM context, never shows up in your logs

speed:

~40ms

check:

before every call

seen by:

nobody

Audit

Every

Black Forest Labs

tool call is logged — who triggered it, which meeting was fetched, what came back. 90 days of history, tied to the user who authorized it

history:

90 days

export:

SIEM-ready

logged:

every call

Test other agents

See the same per-user auth pattern across other AI image and design connectors.

GTM and RevOps Teams

Competitive intelligence briefing agent

Scans Gong calls for competitor mentions, matches each one to its Notion battlecard, and DMs every affected rep a single Slack digest per cycle. Every call runs as the PMM who owns the briefing, never a shared bot.

Why Scalekit

Secure your agent's access. Connectors ship in minutes

01.

Shared tokens break per-user analytics

A shared Black Forest Labs API key looks fine in a demo. In production every generation request looks like one service account, and you cannot tell which user triggered it. Scalekit resolves the credential of the actual user who triggered the agent, never a shared bot.

// shared key
audit → bot_service_account

// scalekit
audit → user_abc ✓

02.

Authentication is not authorization

03.

Multi-tenancy is architectural

04.

Black Forest Labs today. Ten connectors tomorrow.

“Our agents act across Salesforce, Gong, Google Drive, and more, on behalf of every customer. Scalekit behind the scenes meant we can keep adding tools without ever rebuilding how credentials or tool calling work.”

Venu Madhav Kattagoni

Head of Engineering / Von

FAQs

Frequently Asked Questions

Does the agent access Black Forest Labs as the user or through a shared API key?

As the user. Each user connects their own Black Forest Labs API key once, and Scalekit sends it with every call. Audit logs attribute every image generation and edit to that user, not a shared service account.

Where is the Black Forest Labs API key stored?

In Scalekit's managed AES-256 token vault, namespaced per tenant. To rotate it, the user reconnects with a new API key. Revocation is a single dashboard action. Credentials never appear in prompts, logs, or LLM context.

Can I limit what the agent does in Black Forest Labs?

Yes. Filter by tool name in listScopedTools to expose only what you want, for example the image generation tools and blackforestlabs_get_result without blackforestlabs_generate_video and blackforestlabs_upscale_video. Scalekit also enforces scope checks before every API call.

What happens when a user revokes Black Forest Labs access?

The connection is invalidated on the next tool call. Subsequent requests for that user fail closed with a clear error. Other users in the tenant remain unaffected. The event is logged for audit.

How does the agent get the finished image?

Every generation and edit tool returns a task id and polling_url instead of the image. The agent calls blackforestlabs_get_result with the id until the status is Ready, then reads the output. blackforestlabs_get_credits shows the balance before large jobs.