Integration

WaveSpeedAI integration

WaveSpeedAI (wavespeed.ai) runs 1,000+ hosted image, video, audio and 3D generation models behind one API. This integration runs a curated set: text-to-image and image editing (Z-Image Turbo, Seedream 5 Lite, Nano Banana 2, GPT Image 2, Qwen Image 3, FLUX.2 Klein), image upscaling, background and watermark removal, video upscaling, text-to-video and image-to-video (Seedance 2.0, Kling 3.0, Veo 3.1 Fast, WAN 2.6), text-to-speech (ElevenLabs v3, MiniMax Speech 2.6), music (ElevenLabs Music, MiniMax Music 2.5), talking avatars (InfiniteTalk) and image-to-3D (TRELLIS.2). Each model tool submits the job, waits for it and returns the output URLs; every run is charged to the WaveSpeedAI balance, so quote it first with get_model_price and check get_balance. Input media is a public URL; upload_media hosts a local file. list_models, list_predictions, search_billings and get_usage_stats read the catalog and the account's history. Not exposed: the other models (no generic run tool, which could post to any account route), the OpenAI-compatible LLM API, sync and base64 output modes, webhooks and LoRA training. Authentication is an API key (wavespeed.ai/accesskey) sent as a Bearer token.

Connect WaveSpeedAI How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in WaveSpeedAI.Built from the 36 WaveSpeedAI tools below, on the connection you add once.

What you can do with WaveSpeedAI

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Each badge shows the default agent permission of an endpoint (allowed: an agent calls it without asking; ask first: you are asked before an agent calls it, when the agent or chat asks permission for sensitive actions; never: no agent may call it), which every user can change endpoint by endpoint in their credential settings.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Image, 10 operations

flux_2_klein_9b_text_to_imageFlux 2 klein 9b text to imagePOSTAgents: ask firstGenerates an image from a prompt with FLUX.2 [klein] 9B, with crisp text. One flat price per image; outputs holds the image URL. Waits for the job and returns the finished result; gives up after 10 minutes.
gpt_image_2_edit_imageGpt image 2 edit imagePOSTAgents: ask firstEdits one or more images from an instruction with OpenAI GPT Image 2 Edit. Price depends on quality, resolution and the number of images. Waits for the job and returns the finished result; gives up after 10 minutes.
gpt_image_2_text_to_imageGpt image 2 text to imagePOSTAgents: ask firstGenerates an image from a prompt with OpenAI GPT Image 2, good at readable in-image text. Price depends on quality and resolution. Waits for the job and returns the finished result; gives up after 10 minutes.
nano_banana_2_edit_imageNano banana 2 edit imagePOSTAgents: ask firstEdits up to 14 images from an instruction with Google Nano Banana 2 Edit (Gemini 3.1 Flash Image). Price depends on resolution and search options. Waits for the job and returns the finished result; gives up after 10 minutes.
nano_banana_2_text_to_imageNano banana 2 text to imagePOSTAgents: ask firstGenerates an image from a prompt with Google Nano Banana 2 (Gemini 3.1 Flash Image). Price depends on resolution and search options. Waits for the job and returns the finished result; gives up after 10 minutes.
qwen_image_3_text_to_imageQwen image 3 text to imagePOSTAgents: ask firstGenerates an image from a Chinese or English prompt with Alibaba Qwen Image 3.0; 1k and 2k cost the same. outputs holds the image URL. Waits for the job and returns the finished result; gives up after 10 minutes.
seedream_v5_lite_edit_imageSeedream v5 lite edit imagePOSTAgents: ask firstEdits or combines up to 10 reference images from a prompt with Seedream 5.0 Lite Edit. One flat price per image; outputs holds the image URL. Waits for the job and returns the finished result; gives up after 10 minutes.
seedream_v5_lite_text_to_imageSeedream v5 lite text to imagePOSTAgents: ask firstGenerates an image from a prompt with Seedream 5.0 Lite, strong at text in posters. One flat price per image; outputs holds the image URL. Waits for the job and returns the finished result; gives up after 10 minutes.
upscale_imageUpscale imagePOSTAgents: ask firstUpscales an image to 2K, 4K or 8K with more detail and clarity. Billed per run; outputs holds the upscaled image URL. Waits for the job and returns the finished result; gives up after 10 minutes.
z_image_turbo_text_to_imageZ image turbo text to imagePOSTAgents: ask firstGenerates a photorealistic image from a prompt with Z-Image Turbo (6B), optionally guided by a reference image. Billed per image; outputs holds the image URL. Waits for the job and returns the finished result; gives up after 10 minutes.

Video, 9 operations

kling_v3_std_image_to_videoKling v3 std image to videoPOSTAgents: ask firstAnimates an image (optional end frame) into a 3 to 15 s video with Kling 3.0 Standard. Billed per second; sound costs 1.5x. Waits for the job and returns the finished result; gives up after 10 minutes.
kling_v3_std_text_to_videoKling v3 std text to videoPOSTAgents: ask firstGenerates a 3 to 15 s video from a prompt with Kling 3.0 Standard. Billed per second; sound costs 1.5x. Waits for the job and returns the finished result; gives up after 10 minutes.
seedance_2_image_to_videoSeedance 2 image to videoPOSTAgents: ask firstAnimates a start image (optional last frame) into a 4 to 15 s video with native audio with ByteDance Seedance 2.0. Billed per second by resolution. Waits for the job and returns the finished result; gives up after 10 minutes.
seedance_2_text_to_videoSeedance 2 text to videoPOSTAgents: ask firstGenerates a 4 to 15 s video with native audio from a prompt and optional references with ByteDance Seedance 2.0. Billed per second by resolution. Waits for the job and returns the finished result; gives up after 10 minutes.
upscale_videoUpscale videoPOSTAgents: ask firstUpscales a video to 720p, 1080p, 2K or 4K. Billed per second of input (3 s minimum) at the target resolution's rate. Waits for the job and returns the finished result; gives up after 10 minutes.
veo_3_1_fast_image_to_videoVeo 3 1 fast image to videoPOSTAgents: ask firstAnimates an image (optional last frame) into a 4 to 8 s video, with audio by default, with Google Veo 3.1 Fast. Billed per second by resolution and audio. Waits for the job and returns the finished result; gives up after 10 minutes.
veo_3_1_fast_text_to_videoVeo 3 1 fast text to videoPOSTAgents: ask firstGenerates a 4 to 8 s video, with audio by default, from a prompt with Google Veo 3.1 Fast. Billed per second by resolution and audio. Waits for the job and returns the finished result; gives up after 10 minutes.
wan_2_6_image_to_videoWan 2 6 image to videoPOSTAgents: ask firstAnimates an image into a 5, 10 or 15 s video with Alibaba WAN 2.6, optionally guided by audio. Price depends on duration and resolution. Waits for the job and returns the finished result; gives up after 10 minutes.
wan_2_6_text_to_videoWan 2 6 text to videoPOSTAgents: ask firstGenerates a 5, 10 or 15 s video from a prompt with Alibaba WAN 2.6, optionally guided by audio. Price depends on duration and resolution. Waits for the job and returns the finished result; gives up after 10 minutes.

Prediction, 3 operations

delete_predictionsDelete predictionsPOSTAgents: ask firstPermanently deletes the request content and outputs of finished predictions by id; running ones and unknown ids are skipped (see deleted_ids). It does not cancel a run, refund it or remove its billing record.
get_prediction_resultGet prediction resultGETAgents: ask firstReturns one prediction (a model run): its status (created, processing, completed, failed, cancelled, timeout or deleted), output URLs once completed, and the error when it failed.
list_predictionsList predictionsPOSTAgents: ask firstLists a page of your prediction history (API and web runs) with status, model, outputs and timings; filter by status, model, exact id or time window (3 months at most per call).

Music, 2 operations

elevenlabs_music_generate_songElevenlabs music generate songPOSTAgents: ask firstGenerates a song or instrumental (5 s to 5 min) from a description with ElevenLabs Music, billed per second of requested length. Waits for the job and returns the finished result; gives up after 10 minutes.
minimax_music_2_5_generate_songMinimax music 2 5 generate songPOSTAgents: ask firstGenerates a full song with AI vocals from a style prompt and lyrics with MiniMax Music 2.5. One flat price per song; outputs holds the audio URL. Waits for the job and returns the finished result; gives up after 10 minutes.

Speech, 2 operations

eleven_v3_text_to_speechEleven v3 text to speechPOSTAgents: ask firstConverts text (up to 5,000 characters) to speech with ElevenLabs v3, in a preset voice or any ElevenLabs voice id. outputs holds the audio URL. Waits for the job and returns the finished result; gives up after 10 minutes.
minimax_speech_2_6_turbo_text_to_speechMinimax speech 2 6 turbo text to speechPOSTAgents: ask firstConverts text to speech with MiniMax Speech 2.6 Turbo: system or cloned voices, emotion, speed and pitch, 40+ languages. outputs holds the audio URL. Waits for the job and returns the finished result; gives up after 10 minutes.

3d model, 1 operation

trellis_2_image_to_3dTrellis 2 image to 3dPOSTAgents: ask firstTurns one object image into a textured 3D model (GLB) with TRELLIS.2. Price depends on resolution and a 4096 texture; outputs holds the GLB URL. Waits for the job and returns the finished result; gives up after 10 minutes.

Avatar video, 1 operation

infinitetalk_talking_avatar_videoInfinitetalk talking avatar videoPOSTAgents: ask firstTurns a portrait and an audio track into a talking or singing avatar video with InfiniteTalk. Billed per second of audio; 720p costs twice 480p. Waits for the job and returns the finished result; gives up after 10 minutes.

Balance, 1 operation

get_balanceGet balanceGETAgents: ask firstReturns the WaveSpeedAI account balance in US dollars, which every model run is charged to. In a team organization the key must be created by an Owner, Admin, Developer or Billing member.

Billing record, 1 operation

search_billingsSearch billingsPOSTAgents: ask firstSearches the account's billing records (charges per prediction, LLM token usage), newest or oldest first, by type, prediction or order ids, access key, model and time window, with has_more paging.

Image background, 1 operation

remove_image_backgroundRemove image backgroundPOSTAgents: ask firstRemoves the background of an image and returns a transparent PNG cutout (alpha channel). Billed per image; outputs holds the PNG URL. Waits for the job and returns the finished result; gives up after 10 minutes.

Image watermark, 1 operation

remove_image_watermarkRemove image watermarkPOSTAgents: ask firstRemoves text and logo watermarks from an image, keeping texture and background; use it only on images you have the rights to. Billed per run. Waits for the job and returns the finished result; gives up after 10 minutes.

Media, 1 operation

upload_mediaUpload mediaPOSTAgents: ask firstUploads an image, video or audio file (up to 100 MB) to WaveSpeedAI storage and returns its download_url, to pass to a model tool's image, video or audio input. Files are kept 7 days.

Model, 1 operation

list_modelsList modelsGETAgents: ask firstLists every model WaveSpeedAI hosts (1,000+): model_id, name, type (such as text-to-image or image-to-video), description and base price in USD. Use a model_id with get_model_price.

Model price, 1 operation

get_model_priceGet model pricePOSTAgents: ask firstQuotes a model run in USD before you submit it: the base price and your discount, or (with inputs) the price of that exact request, since duration, resolution, audio and other inputs can change it.

Usage statistics, 1 operation

get_usage_statsGet usage statsPOSTAgents: ask firstReturns usage statistics over a window of up to 31 days (default: the last 24 hours): cost and run counts per model, per day, and in total. Organization-wide; needs an Owner, Admin or Billing key in a team.

Use WaveSpeedAI from Claude Code, Cursor or Codex

Connect WaveSpeedAI in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its WaveSpeedAI operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect WaveSpeedAI

Create your workspace
Sign up free. Nothing to install.
Connect WaveSpeedAI once
Add your WaveSpeedAI connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.