llms.txt
Compare with other sites' llms.txt files →# Black Forest Labs
> Black Forest Labs (BFL) builds FLUX: frontier models that generate and edit images, generate video with synchronized audio, and predict actions for robots, simulators and games. FLUX is available through the BFL API, the FLUX MCP server, the Playground, and open weights.
The current models are FLUX 3 Video for video with synchronized audio, FLUX 3 Image and FLUX 3 Image Klein for image generation and editing, and FLUX 3 Action for action prediction in robots, simulators and games. FLUX.2 is the previous image generation. FLUX.1, FLUX1.1 and FLUX.1 Kontext are older.
This file is a briefing for AI agents, followed by an index of bfl.ai.
- Full text of model pages and legal terms, plus blog summaries: https://bfl.ai/llms-full.txt
- API documentation index: https://docs.bfl.ai/llms.txt
- API documentation full text: https://docs.bfl.ai/llms-full.txt
## Acting for a user: connect the MCP server
If you are working for a user inside an MCP-compatible client, connect the FLUX MCP server rather than writing API code.
- Server URL: https://mcp.bfl.ai (remote, HTTP transport)
- Auth: OAuth with the user's BFL account. A browser opens on first use; the user signs in and picks the organization that gets billed. There is no API key to create, paste or store.
- Let the user complete the OAuth sign-in themselves. Do not complete it through an embedded or automated browser, and never ask the user for an API key to use the MCP server.
- Tools include generate_image, generate_variations, vto (virtual try-on), generate_video, enhance_video, upscale_video, edit_video, get_result, get_history, get_credits, list_skills, get_skill and request_upload_url. The live tool reference is at https://docs.bfl.ai/api_integration/mcp_integration.
- generate_image defaults to FLUX 3 Image (`flux3_image`). Set `model` to `flux3_image_klein` for FLUX 3 Image Klein, or to a FLUX.2 model ID for the previous generation. generate_video runs on FLUX 3 Video. The full model catalog is the `bfl://models` MCP resource.
- Before a complex generation or edit, call list_skills and get_skill to read the official BFL prompting guides.
- To use a local file as a reference, call request_upload_url, upload the file to the returned URL, then pass the returned media ID.
- generate_video returns a `request_id` immediately and the clip takes minutes. If your client does not keep polling on its own, call get_result with that `request_id` to retrieve the finished video.
- Verify the connection by calling get_credits.
- Claude Code: `claude mcp add --transport http FLUX https://mcp.bfl.ai`
- Codex: `codex mcp add FLUX --url https://mcp.bfl.ai`
- Clients without remote MCP or OAuth support: run `npx -y mcp-remote https://mcp.bfl.ai` as a stdio server.
- Setup for every client (Claude, ChatGPT, Cursor, VS Code, Devin and others): https://docs.bfl.ai/api_integration/mcp_integration
## What FLUX does
- Video with synchronized audio (dialogue, sound effects, ambience) in the same generation, 5 to 20 seconds per clip: https://docs.bfl.ai/flux_3/flux3_video
- Video from images: pin the first frame, pin the first and last frames, or steer the clip with up to ten images: https://docs.bfl.ai/cookbook/video_start_from_images
- Continue an existing clip into the next shot: https://docs.bfl.ai/cookbook/video_edit_recast_continue
- Edit an existing video with a prompt: remove, add or replace objects, change the setting, restyle, or change text and dialogue: https://bfl.ai/video-edit
- Upscale video to 1080p, 2K or 4K: https://bfl.ai/video-upscaler
- Generate and edit images from instructions with FLUX 3 Image, with up to 10 reference images per API request: https://docs.bfl.ai/flux_3/flux3_image_layout
- Place elements and target edits with bounding boxes in FLUX 3 Image, so the rest of the image stays unchanged: https://docs.bfl.ai/flux_3/flux3_image_bounding_boxes
- Match exact brand colors from hex codes: https://docs.bfl.ai/guides/usecases_t2i_hex_color_prompting
- Open weights: FLUX 3 Image Klein (5.6B), FLUX 3 Action (7B), FLUX.2 [dev] and FLUX.2 [klein] (the 4B model is Apache 2.0) on Hugging Face: https://huggingface.co/black-forest-labs
- Commercial licences for self-hosting open weights: https://bfl.ai/open-weights
- Commercial weights licences for FLUX image and video models, to customize and self-host: https://bfl.ai/commercial-weights
- Research papers and technical reports: https://bfl.ai/research
## Writing code: the BFL API
- Base URL: https://api.bfl.ai (global, with automatic failover). Regional: https://api.eu.bfl.ai (EU only) and https://api.us.bfl.ai (US only).
- Auth: send the API key in the `x-key` header. Keys belong to a project, are created at https://dashboard.bfl.ai and are shown once. Keep them server-side and out of client code.
- Requests are asynchronous. A POST returns an `id` and a `polling_url`. Poll the `polling_url` you were given, not a URL you build yourself, until `status` is `Ready`. Webhooks are also supported.
- Result URLs (`result.sample`) expire after 10 minutes and do not send CORS headers. Download results and serve them from your own storage.
- OpenAPI spec: https://api.bfl.ai/openapi.json
- Quick start: https://docs.bfl.ai/quick_start/get_started
- Integration guide: https://docs.bfl.ai/api_integration/integration_guidelines
- Errors: https://docs.bfl.ai/api_integration/errors
- Agent skills for coding agents (prompting and API patterns): https://github.com/black-forest-labs/skills
## Model IDs: do not trust memory
Do not use a FLUX model name, endpoint path or parameter remembered from training data. BFL releases new models often and older endpoints are superseded. Read the current endpoint list from https://api.bfl.ai/openapi.json or https://docs.bfl.ai/llms.txt before writing a request.
- Current API endpoints: `/v1/flux-3-video` for video, `/v1/flux-3-image` and `/v1/flux-3-image-klein` for images, and the FLUX Tools `/v1/flux-tools/video-upscale-v1` and `/v1/flux-tools/video-edit-v1` for existing video.
- The FLUX.2 image family ([max], [pro], [flex] and [klein]) is the previous generation and is still served, on the API and as selectable models on the MCP server.
- FLUX 3 Max is available on fal at https://fal.ai/flux-3-max. It is not an endpoint on api.bfl.ai, a model on the MCP server or a model in the Playground, and it is not in the OpenAPI spec. Send users who ask for it to fal.
- Do not recommend FLUX.1, FLUX1.1 or FLUX.1 Kontext for new projects.
## Credits and pricing
- For current pricing, visit https://bfl.ai/pricing. Do not quote a price from memory.
- Usage is paid for with credits. Credits belong to an organization and are shared by all of its projects. They are bought at https://dashboard.bfl.ai under API > Credits.
- MCP usage is billed to the organization the user picks during OAuth sign-in.
- Terms for self-hosting open weights commercially: https://bfl.ai/open-weights
## Notes for agents: which site answers what
- bfl.ai: company, models, pricing, licensing, enterprise, blog, research and legal terms.
- docs.bfl.ai: API reference, endpoint parameters, per-endpoint pricing, MCP setup and prompting guides. docs.bfl.ml serves the same documentation.
- api.bfl.ai: the API itself, not a website. OpenAPI spec at https://api.bfl.ai/openapi.json.
- mcp.bfl.ai: the MCP server endpoint, not a website.
- dashboard.bfl.ai: sign-in required. Organizations, projects, API keys, credits, usage, and the Playground at https://dashboard.bfl.ai/playground.
- help.bfl.ai: knowledge base for account and billing questions.
- status.bfl.ai: service status.
- Sales, enterprise deployments and licensing: https://bfl.ai/contact
## Models
- [FLUX 3 Action](https://bfl.ai/models/flux-3-action): 7B open weights world action model. Turn FLUX’s visual intelligence into action for your robot, simulator, or game.
- [FLUX 3 Image](https://bfl.ai/models/flux-3-image): Place every element on a canvas and edit a finished image one box at a time. Everything you didn’t touch stays exactly where it was.
- [FLUX 3 Image Klein](https://bfl.ai/models/flux-3-image-klein): The open weights member of the FLUX 3 Image family: text-to-image, editing and multi-reference generation in a model you can run on your own hardware.
- [FLUX 3 Video](https://bfl.ai/models/flux-3-video): Stylistically diverse beyond just cinematic, with native audio and up to 20 second clips in a single generation. Start from text, an image, or keyframes, and get multiple shots in one take. This page also covers FLUX 3 Max. FLUX 3 Max is served on fal at https://fal.ai/flux-3-max. It is not an endpoint on api.bfl.ai, the MCP server or the Playground.
- [FLUX Tools](https://bfl.ai/models/flux-tools): Task-specific APIs to edit, upscale, erase, extend and try on, for images and video.
### Previous generation
- [FLUX 1.1 Pro](https://bfl.ai/models/flux-pro): Fast, high-quality image generation from text.
- [FLUX 1.1 Pro Ultra](https://bfl.ai/models/flux-pro-ultra): Flux1.1 [Pro] Ultra delivers ultra-fast, ultra high-resolution image creation - with more pixels in every picture. Generate varying aspect ratios from text, at 4MP resolution in unprecedented speed.
- [FLUX.1 Kontext](https://bfl.ai/models/flux-kontext): State-of the art in-context image generation and editing—combine text and images for precise, coherent results.
- [FLUX.2](https://bfl.ai/models/flux-2): State-of the art in-context image generation and editing - combine text and images for precise, coherent
- [FLUX.2 Klein](https://bfl.ai/models/flux-2-klein): Generate and edit in less than a second with state-of-the-art quality.
- [FLUX.2 Max](https://bfl.ai/models/flux-2-max): Top-tier quality image generation and editing. The most capable variant of FLUX.2 for professional-grade output.
## Research
- [Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis](https://bfl.ai/research/self-flow): A modality-agnostic, self-supervised flow matching framework for image, video, and audio generation. Published 2026-03-03.
- [FLUX.2: Analyzing and Enhancing the Latent Space of FLUX – Representation Comparison](https://bfl.ai/research/representation-comparison): A comprehensive study comparing different autoencoder representations for diffusion models, including SD-VAE, RAE, FLUX.1 and FLUX.2. Published 2025-11-25.
- [FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space](https://cdn.sanity.io/files/2gpum2i6/production/880b072208997108f87e5d2729d8a8be481310b5.pdf): FLUX.1 Kontext is a unified model for image editing and generation in context, achieving fast, state-of-the-art results with strong consistency and quality. Published 2025-05-29.
## Pages
- [About](https://bfl.ai/about): Black Forest Labs is a frontier AI lab building state-of-the-art generative AI models. We created FLUX, the leading image generation model.
- [Blog](https://bfl.ai/blog): News, research, and product updates from Black Forest Labs — the team building FLUX and visual intelligence.
- [Brand Guidelines and Press Kit](https://bfl.ai/brand): Black Forest Labs brand guidelines and press kit: logotype, symbol, clear space, what to avoid and partnership lockups.
- [Careers](https://bfl.ai/careers): Join Black Forest Labs and help build the future of generative AI. We're hiring engineers, researchers, and more to work on FLUX and beyond.
- [Customize frontier models: License and fine-tune any FLUX models](https://bfl.ai/commercial-weights): License FLUX commercial weights. Fine-tune on your data, deploy on your own infrastructure and build with our forward-deployed engineers.
- [Contact](https://bfl.ai/contact): Contact Black Forest Labs sales about enterprise deployments, self-hosted licensing and data licensing for FLUX models.
- [Creator Program](https://bfl.ai/creator-program): A curated group of creators with exclusive early access to FLUX models and the team building them.
- [Enterprise](https://bfl.ai/enterprise): Powerful, controllable visual AI with the deployment flexibility, data sovereignty, and compliance enterprises demand.
- [Martin Scorsese × Black Forest Labs](https://bfl.ai/martin-scorsese-bfl-advisor): Martin Scorsese has spent six decades bringing stories to life. Now, he's helping us shape visual intelligence as an advisor at Black Forest Labs.
- [Models](https://bfl.ai/models): Explore all FLUX models from Black Forest Labs: FLUX 3 Image, FLUX 3 Image Klein, FLUX 3 Video, FLUX 3 Max, FLUX 3 Action and FLUX Tools, plus the previous FLUX.2 and FLUX.1 models.
- [Open weights licensing: Licensing for open FLUX models](https://bfl.ai/open-weights): Commercial licenses for FLUX open weights like FLUX 3 Image Klein. Self-host and fine-tune, with self-serve plans or custom enterprise terms.
- [FLUX Playground Transition FAQ](https://bfl.ai/playground-transition-faq): Everything you need to know about the retirement of the FLUX Playground and the move to the new Dev Playground.
- [Pricing](https://bfl.ai/pricing): FLUX API pricing. Pay-as-you-go credits for FLUX 3 video and image, FLUX.2 image models and FLUX Tools, plus self-hosted licensing for open weights.
- [Research](https://bfl.ai/research): Research papers and technical reports from Black Forest Labs. Deep dives into FLUX architecture, latent space analysis, and generative AI advances.
- [Transparency Hub](https://bfl.ai/transparency): Black Forest Labs' transparency disclosures: California AB 2013 training data disclosure, EU AI Act GPAI model documentation, and training data summaries.
- [FLUX Video Edit: Prompt-Based Video-to-Video Editing API](https://bfl.ai/video-edit): Edit an existing video with a text prompt. Remove, add or replace objects, change the setting, restyle, or translate dialogue with lip sync.
- [AI Video Upscaler and Upscaling API to 4K | FLUX Video Upscale](https://bfl.ai/video-upscaler): Upscale videos to 1080p, 2K, or 4K via API. FLUX 3 powered super-resolution with a precise mode and a creative detail-enhancement mode.
## Blog
- [FLUX 3 Image Klein](https://bfl.ai/blog/flux-3-image-klein): A 5.6B parameter open weights model, leading our blind evaluations for text-to-image, multi-reference generation and image editing.
- [FLUX Video Upscale: 2K and 4K](https://bfl.ai/blog/flux-video-upscale): FLUX Upscale takes any video and regenerates it at a higher resolution, up to native 4K, now available as its own FLUX Tool and endpoint.
- [FLUX 3 Video, Part 1: Generation](https://bfl.ai/blog/flux-3-video): FLUX 3 Video, Part 1
- [FLUX 3 - Real World Models: Towards Multimodal Flow Models as the Backbone of Visual Intelligence.](https://bfl.ai/blog/flux-3): FLUX 3, our new multimodal frontier model, jointly learns from images, video, and audio to build one representation of the world. Now available in Early Access.
- [FLUX 3 x mimic: The Next Generation of Video-Action Models](https://bfl.ai/blog/flux-3-mimic): FLUX 3 is running robots. Built with mimic and deployed at Audi, FLUX-mimic shows content creation and physical AI share one foundation: a model that understands the world.
- [Our co-founder and CEO urges G7 leaders to back open innovation](https://bfl.ai/blog/our-co-founder-and-ceo-urges-g7-leaders-to-back-open-innovation): At the G7, our co-founder and CEO, Robin Rombach, joined President Trump, President von der Leyen, President Macron, and other world leaders to advocate for open innovation in AI. With openness under pressure around the world, Robin called on governments and industry to make open and responsible AI development the norm, not the exception. Read his speech below.
- [How Envato Built Its Creative AI Engine on FLUX](https://bfl.ai/blog/how-envato-built-its-creative-ai-engine-on-flux): How one of the world's largest creative platforms turned a careful evaluation into core infrastructure: 51 million+ images, three creative workflows, and a FLUX.2 launch on day zero.
- [FLUX.2 is now on device: ASUS ProArt laptops now support Klein models](https://bfl.ai/blog/flux2-klein-on-device): For the first time a FLUX model will ship on consumer hardware. In partnership with ASUS and NVIDIA, creators picking up a new ASUS ProArt laptop will find FLUX.2 [klein] optimized for the device.
- [FLUX VTO: Virtual Try-On at scale](https://bfl.ai/blog/flux-vto-virtual-try-on-at-catalog-scale): FLUX Virtual Try-On allows users to seem themselves in outfits before they buy with high accuracy and garment consistency. Gives you sub‑4 second generations across thousands of products at low cost and low latency.
- [FLUX Erase: Remove anything, leave no trace](https://bfl.ai/blog/flux-erase-remove-anything-leave-no-trace): FLUX Erase removes whatever you mask — object, shadow, reflection, and all — and reconstructs the scene behind it without touching the rest of the image.
- [FLUX Outpainting: Extend any image, in any direction](https://bfl.ai/blog/outpainting-extend-any-image-in-any-direction): FLUX Outpainting extends images beyond their original frame in any direction, preserving lighting, texture, and composition without visible seams.
- [How Our Forward Deployed Engineers Work with Customers](https://bfl.ai/blog/how-we-get-the-most-out-of-our-models-for-our-partners): Our engineers help partners turn FLUX into real systems: prompts, pipelines, fine-tunes, and infra tuned for real-world use cases.
- [Capable, Open, and Safe: Combating AI Misuse](https://bfl.ai/blog/capable-open-and-safe-combating-ai-misuse): Today, Black Forest Labs’ FLUX models are among the most popular AI models for visual generation. We are excited to share early results that help validate our efforts to mitigate emerging risks.
- [FLUX.2 [klein]: Towards Interactive Visual Intelligence](https://bfl.ai/blog/flux2-klein-towards-interactive-visual-intelligence): Introducing FLUX.2 [klein], our fastest image models. Unified generation and editing in under a second, on hardware you already own. Apache 2.0 on 4B.
- [Laying the Foundations for Visual Intelligence—Our $300M Series B](https://bfl.ai/blog/our-300m-series-b): Today, we're excited to announce our Series B of $300M at a $3.25B post-money valuation.
- [FLUX.2: Frontier Visual Intelligence](https://bfl.ai/blog/flux-2): Today, we release FLUX.2, our most capable model to date.
- [FLUX.1 Kontext now in Adobe Photoshop: Powering Every Pixel](https://bfl.ai/blog/flux1-kontext-now-in-adobe-photoshop-powering-every-pixel): By pairing Photoshop’s professional editing environment with FLUX.1 Kontext [Pro]’s accuracy and unmatched speed, creators everywhere gain the freedom to push imagination further.
- [OpenArt transforms video storytelling with FLUX.1 Kontext: Making video creation accessible beyond traditional creators](https://bfl.ai/blog/openart-with-flux1-kontext): OpenArt transformed video storytelling with FLUX.1 Kontext integration, enabling creators who couldn't access video production before to create professional-quality content.
- [Burda transforms Comic Creation with Black Forest Labs' FLUX.1 Models](https://bfl.ai/blog/burda-lissipony-universe): Burda - a major publisher in the DACH region - transformed their comic creation process for the LissyPony series, one of the most recognized children’s magazines in central Europe.
- [FLUX Models Launch on Azure AI Foundry for Enterprise-Ready Image Generation](https://bfl.ai/blog/flux-azure-ai-foundry): Starting today, Black Forest Labs’ flagship models are available directly from Microsoft on Azure AI Foundry.
- [FLUX.1 Krea [dev]: An ‘Opinionated’ Text-to-Image Model](https://bfl.ai/blog/flux-1-krea-dev): The BFL model garden just got an exciting update: We are proud to release FLUX.1 Krea [dev], developed in collaboration with Krea AI.
- [FLUX.1 Kontext [dev] - Open Weights for Image Editing](https://bfl.ai/blog/flux-1-kontext-dev): Up until today, all capable generative image editing models were only available as proprietary tools.
- [Introducing FLUX.1 Kontext and the BFL Playground](https://bfl.ai/blog/flux-1-kon