public MCP server
AI workflows, paid in Nano over x402.
Image, video, audio, and text pipelines behind one MCP endpoint. No account, no API key — paying is the whole handshake. Built with nanoodle.
claude mcp add --transport http noodles https://mcp.nanoodle.com/mcp
works with Claude Code, Grok, Cursor, and any MCP client
one line registers every workflow with your agent.
your agent picks a workflow and calls it — a poster, a soundtrack, an OG card.
settled automatically in Nano. failed runs are refunded.
No tab, no tip, no signup.
What a call costs
When a tool needs payment, your agent shows you a link with a QR code — scan it with any Nano wallet and the result streams back seconds later. What you pay up front is a deposit: each run settles at the model's metered cost + 20%, and the difference is returned to your wallet on-chain. That 20% is the workflow author's cut: a graph that names a Nano address in its own JSON is paid the markup on-chain, run by run, capped by whatever the deposit has left after the model's cost. None of the workflows here name one yet, so for now the markup stays with this server.
Workflows (14)
Every workflow is a plain noodle-graph.json — open it in the
nanoodle editor to see exactly how it works, remix it, or run it
on your own key. Costs shown are from the last observed run. Each call takes a small deposit, settles at the model's actual cost + 20%, and returns the change.
- Alishan fog dawn
alishan-fog-dawnShareable demo: place + atmosphere → GLM 5.3 Flash writes a 3:2 analog-film still prompt → Muse Image. First-click is Alishan Forest Railway at 5:47am: empty crimson hinoki carriage, cypress cathed…
text→choice→join→llm→image - Character sprites
character-spritesDescription → canonical character → separate cutout parts → local rig and animation baking. One text call + two image calls. The companion rig.mjs extracts transparent parts and bakes 32 frames: id…
text→llm→image→resize→editMade with this skill: Iron Verdict
A playable fighting game built with character-sprites. This graph generates a character reference and parts sheet; the local skill bakes animated sprites, and a coding agent adds combat, gravity and game rules.
- Combine images
combine-imagesUpload two portraits — Muse drops both people into one flooded Vatican-archives heist still.
upload×2→text→editlast run $0.075, ~1m 49sopen in editor · graph JSON - Deslop
deslopPaste AI-sounding text. Grok rewrites it like a person wrote it, a second model grades the rewrite as an AI detector would, then Grok gets the draft, the rewrite and the report and does the final p…
text→llm×2→join×2→llmlast run $0.0341, ~78sopen in editor · graph JSON - Edit a photo
edit-a-photoUpload any photo — Muse rewrites it into one specific typhoon night on Dihua Street, not a hello-world neon filter.
upload→text→editlast run $0.075, ~80sopen in editor · graph JSON - Favicon
faviconDescribe a brand, get a favicon — GLM 5.3 Flash distills it to one bold glyph, Muse Image renders a crisp square. Sample brand is filled so Run works; swap the copy. No uploads. ~$0.02/run (1× GLM …
text×3→join→llm→image - Fibo studio still
fibo-studio-stillFirst-click is Jinguashi at 4:07am: a dented brass carbide miner's lamp on wet black slate, verdigris, sea-fog, one wet quartz cluster. Product + lighting → GLM 5.3 Flash writes a structured photor…
text→choice→join→llm→image - Image model arena
image-model-arenaPick a DIFFERENT model on each Contender, then Run — same Liziba night still four ways, side by side.
text→image×4last run $0.2455, ~1m 39sopen in editor · graph JSON - Night market postcard
night-market-postcardShareable demo: place + vibe → GLM 5.3 Flash writes a 3:2 film-stock postcard prompt → Muse Image. Edit the place, pick a vibe, hit Run. No uploads. Cheap models (1x GLM Flash + Muse Image at 3:2).
text→choice→join→llm→image - Omni flash turntable
omni-flash-turntableShareable demo: object + move → GLM 5.3 Flash writes a cinematic t2v prompt → Gemini Omni Flash 1.1. First-click is a barnacled brass diving helmet turning in dock fog (not a clean studio bottle). …
text→choice→join→llm→tvideo - Photo to video
photo-to-videoStill from a prompt, then MiniMax H3 Spicy (open-weights i2v) animates it — expressive motion, optional last frame, 480p, ~5s. First-click is a woman standing in a red silk dress on a rainy Tokyo f…
text×3→image→llm→ivideo - Render a mockup
render-a-mockupDescribe a screen, keep the style guide: an LLM turns both into one dense image prompt and an image model renders the mockup.
text×2→join→llm→imagelast run $0.079, ~2m 16sopen in editor · graph JSON - Sing
singOne idea fans out to three LLMs — lyrics, a matching style, and a negative prompt of what to avoid — then a music model sings it. First-click is a Fushimi Inari 3:17am torch song: the rusted snack …
text×3→llm→join×2→llm→join→llm→musiclast run $0.0497, ~1m 56sopen in editor · graph JSON - Talking avatar
talking-avatarMuse paints a Hongyadong cliff-walk courier, TTS speaks the last-drop line, Longcat lipsyncs it. Edit the look or the line, hit Run (~$0.15–0.25 at 480p). Replace the Face image node with an Upload…
text×3→image→tts→lipsync
Private by design
Every line here is checkable in the source — nothing to take on faith.
- No accounts, no API keys, no sign-ins — paying is the only thing that authorizes a call.
- Your prompts and inputs are never written to disk or logs.
- Input media rides in memory only — it is never stored.
- Generated media auto-deletes after 24 hours, served at unguessable URLs until then.
- Results are held in memory for delivery, not stored.
- No analytics, no trackers, no IP logs — the server keeps no request logs.
- The only thing written down is a payments ledger, for money events already public on the Nano ledger. No run telemetry, no error text.
One caveat: generation runs on NanoGPT's API, so your prompt content in flight falls under their privacy policy.
Claim the 20% on your own workflow
A graph that declares a Nano address ("x402": {"author": "nano_…"} in its
JSON) has the 20% markup of every paid run sent to it on-chain, automatically — as much of it as the
deposit leaves after the model's cost. This server takes no cut of that markup, and Nano charges no
fee to move it. No workflow here claims it yet — the field is unset on all 14, so the markup stays with this server.
The public library behind this server is awesome-noodles —
add your workflow there with your address to get listed and earn on every run.
Open source — host your own
MIT-licensed: the server, the editor, and the executor. The workflows above are a separate matter — you can grab any graph JSON, but each graph belongs to whoever wrote it and carries whatever licence they give it. One runtime dependency is not MIT either: nanocurrency is GPL-3.0, and it signs the Nano blocks for the optional x402 wallet (details). One command turns your own folder of graphs into a server exactly like this one:
npx nanoodle-mcp --graphs ./noodles --serve 8402
Add --charge-usd 0.05 --public-url https://your-host and a Nano wallet to
charge per call — see the README.