gemini-mcp

MCP server for Google Gemini media generation. Exposes eleven tools to Claude over stdio: list available models, generate/edit/compose images, generate a consistent set of images from a master prompt, multi-turn image refinement (Interactions API), video generation (omni), music generation (Lyria), an async result poll for long generations, and Files API upload/list/delete for reusable image references. Output is written to disk by default (path returned) or returned inline as base64. Built on the Gemini v1beta API (generativelanguage.googleapis.com) using the Nano Banana / Nano Banana Pro (images), omni (video), and Lyria (music) model families.

Developed and maintained by AI (Claude Code).

Environment Variables

Variable	Required	Description
GEMINI_API_KEY	Yes	Your Google Gemini API key (aistudio.google.com/apikey)
GEMINI_IMAGE_MODEL	No	Override the default image model (default: gemini-3.1-flash-image)
GEMINI_OUTPUT_DIR	No	Default directory for generated images (default: current working directory)
GEMINI_INPUT_DIR	No	Directory to resolve bare input-image filenames against (so images: ["foo.jpg"] works)
GEMINI_TIMEOUT_MS	No	Upstream request timeout in ms (default: 60000, or 120000 for image_size: "4K"); each generation tool also takes a per-call timeout_ms
GEMINI_HEARTBEAT_MS	No	Progress-notification cadence in ms while a generation runs (default: 10000; 0 disables) — keeps MCP hosts that reset their timeout on progress from timing out long generations
GEMINI_CHAIN_RETRY_MS	No	How long to wait out interactions-store lag when a chained call 404s (default: 120000; 0 disables retrying)

Long generations and client timeouts

4K / Pro-model generations can outrun an MCP host's own tools/call timeout (error -32001). The server sends notifications/progress heartbeats so hosts that reset their timeout on progress wait it out. If the host still gives up, the server-side generation usually completes anyway: the image is written to the output dir, gemini_interact also writes an <image>.json sidecar recording the interaction_id, and continue_last: true resumes the interaction the lost response belonged to.

When a chained call 404s

A 404 on a request carrying previous_interaction_id is not proof the chain expired. The only 404 body observed live is generic — "Requested entity was not found." — and never names which entity. An unknown or renamed model id, and an expired Files API files/… uri (~48h TTL), return exactly the same thing. So the server no longer asserts a cause it can't establish: the upstream text is surfaced verbatim, and gemini_interact runs an experiment to find out which it was.

Most often the id isn't missing at all — it just isn't visible yet. The interactions store is eventually consistent, and a freshly created id can 404 while the same id resolves fine minutes later; heavy turns (4K, Pro, thinking_level: high) are the likeliest to hit it, which is exactly the turn you most want to chain from. So a chained 404 is retried with exponential backoff for up to 120s (GEMINI_CHAIN_RETRY_MS) before anything is declared broken. The 404 generates nothing and isn't billed, so the wait costs only time.

After that budget is spent, the tool looks up that id's sidecar, re-attaches the image it produced, and re-issues the request without the chain:

• The re-issue succeeds → the chain really was the problem, and you get your image anyway, reported as chain_recovered: { expired_interaction_id, reanchored_on }. The 404'd attempt generates nothing, so this costs the one generation you'd have paid for re-anchoring manually.
• The re-issue 404s too → the interaction id was never the cause. You get told exactly that, with the upstream text, and pointed at the model id and any files/… uri instead of being sent to chase an interaction that was fine all along.
• No sidecar matches the dead id → the original error, rather than a guess. Re-anchoring on the wrong picture would silently corrupt the edit.

Separately, continue_last no longer dies with the server process: with no in-memory id it resumes from the newest <image>.json sidecar in the output dir and reports continued_from_sidecar: true. That case was never an expired chain at all — the interaction was alive upstream the whole time; only our memory of its id was gone.

For hosts whose timeout can't be tamed (e.g. Claude Desktop, a fixed ~30s cap that ignores progress), two guards make re-issuing safe and unnecessary:

• async: true returns a job_id immediately instead of the image, so the call can't time out at all; poll gemini_get_result with the job_id until it's done.
• idempotency_key makes a repeat call idempotent — a retry with the same key returns the recorded result (reused: true) instead of billing a second generation. (Even without a key, two identical in-flight calls are deduplicated automatically.)

Tools

Tool	Description
gemini_list_models	List available Gemini image models and the current default
gemini_image_generate	Generate image(s) from a text prompt
gemini_image_edit	One-off edits or multi-image composition with a text instruction (for a series of edits, use gemini_interact)
gemini_image_set	Generate a master image plus N consistent images referencing it
gemini_interact	Preferred tool for iterative refinement: multi-turn generation/editing via the Interactions API — chain the returned interaction_id via previous_interaction_id (or continue_last: true)
gemini_video_generate	Generate a short video (text→video, image→video, or edit) via the Gemini omni model (preview); written to disk as MP4
gemini_music_generate	Generate music from a text prompt via a Lyria model — lyria-3-clip-preview (~30s, default) or lyria-3-pro-preview (longer, WAV-capable); written to disk as MP3/WAV (preview)
gemini_get_result	Fetch an async generation started with async: true by its job_id (status running → done result). Lets a long generation outlive a host's tools/call timeout
gemini_upload_file	Upload an image (or video/audio) to the Gemini Files API once — from a url, data_base64, or a local path — and get a reusable files/<id> reference
gemini_list_files	List the files currently uploaded under this API key, with MIME types and expiry times
gemini_delete_file	Delete an uploaded file before its ~48h expiry (confirm-gated)
gemini_sign_media	(hosted connector only) Mint a fresh signed URL for generated media from its r2_key — an expired link is not a dead end

Seeing your images (hosted connector)

On the hosted connector there is no filesystem, so a generated image has to come back as something you can open. It does: every result includes a URL, with no configuration.

{
  "images": ["https://connector.example.com/media/gen/2026-07-29/ab12cd34-a-cat.png?exp=…&sig=…"],
  "media":  [{ "url": "https://…", "r2_key": "gen/2026-07-29/ab12cd34-a-cat.png",
               "expires_at": "2026-07-31T12:00:00.000Z",
               "curl_hint": "curl -sS -o a-cat.png \"https://…\"" }]
}

Those links need no auth header — the signature is in the URL — so they work in a browser, in curl, and in a chat message. They expire (48h by default) and the objects behind them are swept on a retention schedule. The r2_key is the durable handle for that window:

• gemini_sign_media (hosted connector only) mints a fresh signed URL from an r2_key, so an expired link never forces you to re-generate — and re-pay for — the image.
• gemini_upload_file with r2_key turns media the connector generated into a Files API reference (the connector reads its own bucket — no HTTP round trip, no signature needed), so a generated image can become the reference image for the next generation in one cheap call.
• Idempotent replays (idempotency_key) re-mint the URLs inside the recorded result before returning it, so a reused result never carries a dead link.

Why this matters: MCP's inline image content blocks (inline: true) are visible to the assistant but many chat clients never render them to the user, and the assistant cannot extract bytes back out of its own context to save them elsewhere. A generation could bill successfully and be invisible. A URL is the portable answer; inline remains available, but it is no longer the only way to receive media.

Three ways to serve the bytes

	setup	URL shape
Worker route (default)	none — works out of the box	https://<worker-host>/media/<key>?exp=&sig=
r2.dev public bucket	enable public access on the bucket, set MEDIA_PUBLIC_BASE_URL	https://pub-….r2.dev/<key>
Custom domain (best)	attach a domain to the bucket in Cloudflare, set MEDIA_PUBLIC_BASE_URL	https://media.example.com/<key>

The built-in Worker route is the zero-config path: the connector streams the object out of R2 itself. The other two are worth setting up if you want plain, non-expiring URLs or want media traffic served by R2 rather than by the Worker — r2.dev is easiest but rate-limited and fine for personal use; a custom domain is the best of the three.

Auth on /media is a signed, expiring URL rather than an unlisted key. Random keys would be simpler, but they never expire and never revoke: anything that ever logged or forwarded the link keeps working forever. A signature scopes access to one object with a deadline, and rotating MEDIA_URL_SECRET invalidates every outstanding link at once. The tradeoff is that links are long and cannot be shortened by hand.

For assistants relaying a result

Show the user the URL. If your sandbox has network egress to the connector host, fetching it and attaching the bytes as a file gives the nicest result; otherwise present the link itself. Whether a given client renders ![](url) markdown inline varies by client — a bare URL is the safe form, and a markdown link is a reasonable enhancement where you know it renders.

Retention

Variable	Default	Effect
MEDIA_TTL_DAYS	7	Objects older than this are deleted by a daily cron (both the current gen/ prefix and the legacy media/ one)
MEDIA_URL_SECRET	generated	HMAC key for /media links; rotate to revoke all outstanding URLs
MEDIA_PUBLIC_BASE_URL	unset	Serve from R2 directly instead of the Worker route

Signed-URL lifetime is clamped to MEDIA_TTL_DAYS, so a link never outlives the object it points at.

Sending reference images without burning context

Every tool that takes a reference image — gemini_image_generate, gemini_image_edit, gemini_image_set, gemini_interact, plus gemini_video_generate (reference stills) and gemini_music_generate — accepts them four ways. Only one of them costs model context:

Parameter	Where the bytes travel	Context cost
images_url (master_images_url)	the server downloads the https URL	none
images_file_uris (master_images_file_uris)	a files/<id> reference, already uploaded	none
images	read off local disk (stdio builds only)	none
images_base64	through the tool-call JSON	~14k tokens per JPEG

images_base64 is the fallback of last resort. It costs roughly 14k tokens per modest photo, and it is silently corrupted whenever the file read that produced it was truncated — the payload still looks like base64, so the failure surfaces as a bad generation rather than an error. Prefer any of the other three.

images_url — the server fetches it

{ "prompt": "make it look like winter", "images_url": ["https://example.com/photo.jpg"] }

Fetches are restricted to public https:// URLs — private, loopback and link-local hosts are refused (IPv6 literals are parsed, so [::ffff:7f00:1] is caught as loopback), every redirect hop is revalidated, and each hop is bounded by a timeout. The response must be Content-Type: image/* and is capped at 15MB, enforced while streaming rather than trusted from Content-Length. A failure names the offending URL. Anything over 6MB is uploaded to the Files API and referenced by uri instead of inlined, since generateContent caps a whole request near 20MB.

images_file_uris — upload once, reference many times

// 1. upload
{ "tool": "gemini_upload_file", "url": "https://example.com/photo.jpg" }
// → { "file_uri": "files/abc123", "mime_type": "image/jpeg", "expires": "..." }

// 2. reference it, as many times as you like
{ "prompt": "make it winter",  "images_file_uris": ["files/abc123"] }
{ "prompt": "make it sunrise", "images_file_uris": ["files/abc123"] }

Uploads are retained ~48h; after that the reference stops resolving (as a generic 404 — see the chained-404 section above). gemini_image_set fetches or resolves such a reference once and passes it to the master and every scene call.

On stdio builds, a local images path that gets referenced more than once in a session is uploaded to the Files API automatically (keyed on path + mtime + size), so repeated edits of the same photo stop re-sending the bytes. Editing the file invalidates the cached upload.

POST /upload — raw bytes, no base64 at all (hosted connector)

The hosted connector exposes an HTTP upload endpoint behind the same OAuth token as /mcp. This is the intended path for an agent with a shell: disk file → curl → file_uri → tool call, with the image never entering the conversation.

curl -X POST https://connector.gemini.nullnet.app/upload \
  -H "Authorization: Bearer $ACCESS_TOKEN" \
  -H "Content-Type: image/jpeg" \
  --data-binary @photo.jpg
# → {"file_uri":"files/abc123","mime_type":"image/jpeg","expires":"...","...":"..."}

Then pass "images_file_uris": ["files/abc123"] to any image tool. The body is streamed straight to the Files API (nothing is buffered), so Content-Length is required — curl sets it automatically with --data-binary @file. Accepted content types are image/*, video/* and audio/*; an optional ?name= or X-Filename header sets the display name.

Quick Start

{
  "mcpServers": {
    "gemini": {
      "command": "npx",
      "args": ["-y", "@chrischall/gemini-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

See SKILL.md for full usage documentation.