Opening mcp.
0.0s
0.0s
Developers · MCP server
One Model Context Protocol endpoint gives Claude, ChatGPT, OpenCode, Cursor, VS Code, Codex and any other MCP client the whole studio: images, video, music, voice, vectors and 3D, on your own prepaid tokens.
https://imgica.ai/mcpClaude does not make images on its own. Connect the Imgica MCP server and Claude can call image, video, music and voice tools directly, using models such as GPT Image 2, Nano Banana Pro, Seedream, Kling and Veo. Sign in once, approve access, then ask Claude for what you want. The same server works in ChatGPT, Cursor and VS Code.
How it connects
Paste https://imgica.ai/mcp into your client. Every client below has a snippet.
Your client opens Imgica, you approve, and it holds a token for your account. Or paste a personal API key from Studio settings, Developer.
The model lists what Imgica can make, prices it in tokens before spending, then creates and waits for the result.
Every result is a URL on Imgica's storage. Pass it straight into the next tool, or download it.
Connect your client
Clients that sign in with OAuth need nothing else. Where a key is shown, make one at Studio settings, Developer, and it drops into the snippet there.
Sign in with OAuth from inside Claude Code, or pass a key header for scripts and CI. Sign-in: OAuth or API key.
claude mcp add --transport http imgica https://imgica.ai/mcp # then type /mcp in Claude Code and sign in # with a key instead: claude mcp add --transport http imgica https://imgica.ai/mcp \ --header "Authorization: Bearer imk_live_YOUR_KEY"
The snippet uses imk_live_YOUR_KEY as a placeholder. Signed-in users see their own key filled in at Studio settings, Developer.
What the model can call
Every create tool prices itself the way the studio does, charges once, and hands back a job id. The model waits, then uses the URL. Failed jobs refund themselves.
models:read Browse models, prices, voices and stylesaccount:read Read your plan and token balancemedia:read Read your jobs and their outputsmedia:write Upload input files and delete outputsgenerate Create images, video, audio, vectors and 3D (spends your tokens)Discover, no sign-in
list_modelsEvery model of a kind with prices and capabilitiesget_modelOne model's exact parameters, sizes, durations and pricesestimate_costThe tokens a call would charge, from the same pricing that chargeslist_voicesVoices for text to speechlist_filter_stylesStyle presets for the photo filterCreate, spends tokens
generate_imageText to image and image to image on any image modeledit_imageInpaint, relight, recolor, restyle, expand, camera anglesblend_imagesCombine several references with a promptgenerate_videoText to video, image to video, extend, edit, upscale, lip sync, avatarsupscale_mediaImages to 2K, 4K or 8K, video to 1080p, 2K or 4Kremove_backgroundTransparent cutouts for images, clean plates for videogenerate_vectorSVG from a promptapply_filterA style preset or your own look on a photoswap_faceOne face onto another imagegenerate_3dA GLB model from a prompt or an imagegenerate_musicSongs with sung lyrics, or instrumentalstext_to_speechNarration in a chosen voiceenhance_promptA rough idea rewritten into a detailed promptJobs and files
wait_for_jobLong-poll a job until it finishesget_jobThe state and outputs of a joblist_jobsYour library, one kind at a timedelete_jobRemove something from your libraryupload_mediaBring in a file by URL or base64get_accountPlan, balance and limitssearchSearch your library and the catalog (ChatGPT connectors)fetchThe full record behind a search resultQuestions
No. Browsing models and pricing is free and needs no sign-in. Creating something spends tokens from your Imgica account, the same tokens the studio uses. Free tokens at signup, no card required, pay once when you are ready.
Anything that speaks the Model Context Protocol over HTTP: Claude on the web, desktop and mobile, Claude Code, ChatGPT, OpenCode, Cursor, VS Code, Codex, Gemini CLI, and the OpenAI and Anthropic APIs. A client that only speaks stdio bridges through mcp-remote.
Through OAuth. Your client discovers Imgica's authorization server, sends you to an Imgica page to approve, and receives a token bound to your account. Personal API keys exist for scripts and headless use, and you can revoke either from Studio settings, Developer.
Images usually finish within a minute, video within a few minutes, music and speech in between. Create tools return a job id right away; wait_for_job polls for you and can be called again if time runs out.
The tokens come back to your balance automatically, and the job says why it failed. Retrying a create tool with the same idempotencyKey never charges twice.
Yes. Revoke a key or disconnect an app from Studio settings, Developer, and that client stops the moment it next calls.