Image generation and editing APIs and MCP servers for AI agents
Text-to-image, image editing, inpainting, upscaling and background removal.
10 image generation and editing APIs an AI agent can call: auth method and header, official MCP server (5), free tier (2), docs links. Machine-readable JSON included. Agents: fetch catalog/image-generation.json instead of parsing this page. Last checked 2026-09-24.
Ratings and usage (★) fetched: .
| Service | What an agent can do | Auth | MCP server | Free tier | Details (with sources) | Notes | Proof (ratings, usage) |
|---|---|---|---|---|---|---|---|
| Adobe Firefly Services API | Generate and edit images and manage creative assets programmatically via Adobe's Firefly REST APIs. | OAuth 2.0OAuth2 Server-to-Server (Client ID/Secret via Adobe IMS); Authorization: Bearer <token>, token valid 24h | community | check pricing; requires Adobe Developer Console project and Firefly Services entitlement | Base URL: https://firefly-api.adobe.io/v3 Free plan: none ↗ Rate limits: Default 4 requests/minute and 9,000 requests/day per organization; 429 when exceeded ↗ Async jobs: POST /v3/images/generate-async returns jobId, statusUrl, cancelUrl; poll GET /v3/status/{jobId} ↗ OAuth scopes: openid, AdobeID, session, additional_info, read_organizations, firefly_api, ff_apis ↗ MCP tools: 8: firefly_generate_image, firefly_generate_similar, firefly_expand_image, firefly_fill_image, firefly_generate_object_composite, firefly_generate_video … ↗ | Enterprise-leaning; needs OAuth Server-to-Server credentials from Adobe Developer Console, not a simple sign-up API key. | — |
| Black Forest Labs FLUX API | Generate, edit, and vary FLUX model images from text prompts via REST or the official OAuth MCP server. | API keyx-key: <BFL_API_KEY> | official | paid only, no free tier (FLUX Schnell weights are open for self-hosting) | Base URL: https://api.bfl.ai/v1 Free plan: none ↗ Rate limits: Max 24 concurrent requests for most endpoints (6 for flux-kontext-max); 429 when exceeded ↗ Data policy: Generated result URLs on delivery.*.bfl.ai expire after 10 minutes; download promptly ↗ Async jobs: POST returns id + polling_url; poll polling_url (or use webhook) until Ready ↗ Example call: generate an image from a text prompt (FLUX.2 [pro]): POST https://api.bfl.ai/v1/flux-2-pro-preview ↗ Unit price: FLUX.2 [pro] text-to-image from $0.03 per image (megapixel-based, scales with resolution); FLUX1.1 [pro] 4 credits = $0.04 per image; 1 credit = $0.01 ↗ MCP tools: 7: generate_image, vto, generate_variations, get_history, get_credits, generate_video … ↗ | Official MCP server (flux-mcp) is remote/hosted at mcp.bfl.ai with OAuth-only login, billed to your BFL org. | — |
| Freepik API | Generate and edit images, and generate video, music and sound effects via REST API, plus access the Freepik stock asset library. | API keyx-magnific-api-key header (legacy x-freepik-api-key); base URL https://api.magnific.com/v1 | official | free tier: ~5 EUR of free API credits, then pay-as-you-go | Base URL: https://api.magnific.com/v1 Rate limits: Per IP: 50 hits/s over 5 s, and 10 hits/s averaged over 2 min; no daily cap for most endpoints ↗ Async jobs: POST returns task id; poll GET /v1/ai/<model>/{task-id} or pass webhook_url ↗ Example call: generate a video from an image (Kling 2.1 Pro): POST https://api.magnific.com/v1/ai/image-to-video/kling-v2-1-pro ↗ MCP tools: 9: images_generate, images_upscale, images_remove_background, images_to_svg, video_generate, audio_tts … ↗ | Freepik rebranded to Magnific in April 2026; docs/domains and auth header are mid-migration from freepik.com to magnific.com. The official remote MCP uses OAuth sign-in, no API key. | — |
| Google Gemini / Imagen image API | Generate and multi-turn edit images from text via REST (Nano Banana/Imagen models), up to 4K with reference-image support. | API keyx-goog-api-key: $GEMINI_API_KEY | — | Paid only: pricing page lists Free Tier 'Not available' for all Gemini image models (Nano Banana family); usage-billed | Base URL: https://generativelanguage.googleapis.com/v1beta Free plan: none, watermarked ↗ Data policy: Paid tier: content not used to improve Google products; all generated images carry a SynthID watermark ↗ | Also available via Vertex AI for enterprise/cloud_iam auth as an alternative. Google's experimental mcp-genmedia servers (not officially supported) use Vertex AI ADC auth, not Gemini API keys. | — |
| Ideogram API | Generate, edit, remix, upscale, and describe images via REST or an official OAuth MCP server, strong at in-image text. | API keyApi-Key: <IDEOGRAM_API_KEY> | official | Paid only: no free API credits (web app has 10 slow credits/week); API is pay-per-image from prepaid credit (min one-time purchase $1, min auto top-up target $10) | Base URL: https://api.ideogram.ai Free plan: none ↗ Rate limits: Default 10 in-flight requests; contact Ideogram for more ↗ Data policy: Image links returned by the API expire; download and store images you want to keep ↗ MCP tools: 9: generate_images_bulk, edit_image, reframe_image, upscale_image, create_collection, upload_image … ↗ | API billing is separate from Ideogram web subscriptions; requires payment method for API access. The official MCP uses OAuth and draws from the Ideogram web subscription, not API credits. | — |
| Leonardo.Ai API | Generate images and videos from text via REST or the official MCP server, with model selection and job status polling. | API keyAuthorization: Bearer <LEONARDO_API_KEY> | official | Paid only: Pay-As-You-Go API credit (USD balance) via manual or auto top-ups; API access is separate from free/web app plans | Base URL: https://cloud.leonardo.ai/api/rest/v1 Rate limits: Default concurrency 10 generation jobs; extra requests queue up to a pending limit; separate API request rate limit ↗ Async jobs: POST returns generationId; poll GET /generations/{id} or configure webhook callback; a sync endpoint exists for some models ↗ MCP tools: 1: generate-image ↗ | API credit is bought separately from web app subscriptions. The official MCP (https://mcp.leonardo.ai/v1/mcp) authenticates with an 'API-Key' header, not OAuth. | SourceForge 5.0★/1 review |
| OpenAI Images API (gpt-image) | Generate, edit, and iteratively refine images from text prompts via REST, with mask-based inpainting and streaming previews. | API keyAuthorization: Bearer <OPENAI_API_KEY> | community | paid only; no free image credits, usage-based billing | Base URL: https://api.openai.com/v1 Data policy: API data not used for training unless opted in; /v1/images/generations keeps abuse-monitoring data 30 days ↗ Example call: generate an image from a text prompt: POST https://api.openai.com/v1/images/generations ↗ MCP tools: 2: create-image, edit-image ↗ | Requires an OpenAI API account with billing enabled; part of the Images API / Responses API. Current models: gpt-image-2.5-sunburst and gpt-image-2.5-flare; API Organization Verification may be required. | GitHub 111★ SureScaleAI/openai-gpt-image-mcp |
| Recraft API | Generate, edit, upscale, and vectorize images via REST API using Recraft's proprietary models. | API keyAuthorization: Bearer RECRAFT_API_TOKEN | official | paid only; pay-per-image (from $0.007/image on V4.1 Flash), no confirmed free credits | Base URL: https://external.api.recraft.ai/v1 Free plan: none ↗ Rate limits: 100 images/minute per user and 5 requests/second ↗ Data policy: Generated images stored approx. 24 hours (may change) ↗ MCP tools: 9: generate_image, create_style, vectorize_image, image_to_image, remove_background, replace_background … ↗ | REST API, OpenAI-Python-library compatible for some endpoints (not all params supported). Official MCP: hosted at https://mcp.recraft.ai/mcp (OAuth 2.0, bills subscription credits) and a local server on GitHub (recraft-ai/mcp-recraft-server). | GitHub 60★ recraft-ai/mcp-recraft-server |
| remove.bg API | Remove image backgrounds programmatically via a simple REST API with JSON or multipart uploads. | API keyX-Api-Key header | — | Free: first 50 API calls per month, then paid credits | Base URL: https://api.remove.bg/v1.0 Free plan: 50 API calls, per month ↗ Rate limits: Up to 500 images/minute, scaled down by megapixels (e.g. 10/min at 50 MP); 429 with Retry-After ↗ Example call: remove the background from an image: POST https://api.remove.bg/v1.0/removebg ↗ Unit price: 1 credit per full-resolution image (preview 0.25 credit); pay-as-you-go 10 credits for $9 down to 8,000 for $1,699; subscriptions from $9/mo for 40 credits ↗ | IMPORTANT: remove.bg API is being discontinued - stops accepting requests on December 1, 2026, with removal moving to Leonardo.ai. | — |
| Stability AI Platform API | Generate, upscale, and edit images via REST (Stable Image Ultra/Core/SD3 models) using simple credit-based billing. | API keyAuthorization: Bearer <STABILITY_API_KEY> | community | 25 free credits on signup, then pay-as-you-go | Base URL: https://api.stability.ai Free plan: 25 free credits, one-time ↗ Rate limits: 150 requests every 10 seconds; 429 when exceeded ↗ Example call: generate an image from text (Stable Image Core): POST https://api.stability.ai/v2beta/stable-image/generate/core ↗ Unit price: Stable Image Core: 3 credits per successful image; 1 credit = $0.01, so $0.03 per image; failed generations free ↗ MCP tools: 12: generate-image, generate-image-sd35, remove-background, outpaint, search-and-replace, upscale-fast … ↗ | — |
Other categories
- Web search
- Web scraping and browser automation
- Knowledge and research data
- Video generation and editing
- Speech and audio
- Music generation
- 3D generation and assets
- Architecture, CAD and BIM
- Diagrams and software architecture
- Design and UI
- Documents, OCR and presentations
- Translation and language
- Code execution sandboxes
- Developer platforms and DevOps
- Cloud and hosting
- Databases
- Vector search, memory and RAG
- Email, team chat and notifications
- SMS and messaging apps
- Voice agents and telephony
- Productivity and workspace
- CRM, support and marketing
- Social media
- Maps, geolocation and weather
- Finance, payments and market data
- E-commerce
- Automation and integration platforms
- Model APIs and inference