Troquel MCP

Use Troquel from the assistant you already have open

Connect Claude, ChatGPT, Cursor or any MCP client and ask it to analyse a brand, write the prompts or generate a piece. It runs on your credits and everything lands in your library.

Works withclaude.aiClaude DesktopClaude CodeChatGPTCursorVS CodeCodexHermes AgentDeepSeek HarnessGemini CLIGrok

Connect my assistant

Connection guide

Connect your assistant

Choose your AI client and follow its steps. Connect with your Troquel account or an API key, depending on the client. Both use the same balance.

MCP server URL

https://troquel.ai/api/mcp

claude.ai

With your account, no key

Official guide

Free, Pro, Max, Team and Enterprise plans; Free allows one custom connector. On Team and Enterprise an Owner adds it in Organization settings → Connectors, then each member taps Connect.

Open Customize → Connectors in claude.ai.
Open Connectors
Tap + Add, then Add custom connector.
3
Name it Troquel, paste this as the remote MCP server URL and tap Continue:
MCP server URL
```
https://troquel.ai/api/mcp
```
Keep the detected options (Sign in now and Register automatically) and tap Add.
Tap Connect. Troquel opens: sign in and tap Allow.
In a chat, turn it on from the + button → Connectors.

Step 1/7 · Connectors → + Add

For agents without a browser

Scripts, servers or agents that cannot open a browser connect with an API key. The key spends your balance: do not paste it anywhere shared.

Create a key in Settings → API keys and replace tk_YOUR_KEY.

When adding the connector, under Authentication pick No sign-in.
2
Under Request headers add a header named Authorization with this value:
Header value
```
Bearer tk_YOUR_KEY
```
It is in beta: if you do not see Request headers, your account does not have it yet, so connect with your account as above.

Try it: ask «Use Troquel to check my balance».

The apps you connected are listed, and disconnected, in Settings → Connected apps.

API from the terminal

The same API the web uses. You create a job, check its status and download the assets. Good for scripts, n8n or whatever you build.

1Create a job with POST /api/jobs: model, variant, prompt and aspect ratio.
2Save the id returned in the response.
3Poll GET /api/jobs/JOB_ID until status is done: the response includes the result URLs.

Create a job

curl -X POST https://troquel.ai/api/jobs \
  -H "Authorization: Bearer tk_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"modelId":"gpt-image-2","variantId":"1k","prompt":"Cream packshot on marble, window light","aspect":"1:1"}'

Check the result

curl https://troquel.ai/api/jobs/JOB_ID -H "Authorization: Bearer tk_YOUR_KEY"

A 10-second video

Set the duration with duration_seconds and resolution: Troquel selects the matching model variant (for example, Seedance 2.0 at 10 s uses 720p-10). If that combination is unavailable, the error lists the supported durations and resolutions.

From the chat (MCP)

generate_video({
  model: "seedance-2.0",
  duration_seconds: 10,
  resolution: "720p",
  prompt: "Bottle turning on marble, window light",
  aspect: "9:16"
})

Via API (curl)

curl -X POST https://troquel.ai/api/v1/generate \
  -H "Authorization: Bearer tk_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"kind":"video","model":"seedance-2.0","duration_seconds":10,"resolution":"720p","prompt":"Bottle turning on marble, window light","aspect":"9:16"}'

Tools it exposes

What your assistant can call. Each one costs the same as it does on the web.

balance

Current balance in credits.

list_models

Model catalog with variants and prices.

list_brands

Your saved brands, so you can choose the context for each generation.

generate_image

An image with any model, with optional references.

generate_video

Text or image to video. Available durations depend on the model.

Without model: seedance-2.5 (Seedance 2.5 · 480p · 5 s (draft)), 85 credits.

generate_audio

Voice over with ElevenLabs Eleven v4, Argentine voices.

generate_sound_effect

A sound effect from a text, 0.5 to 30 seconds.

add_video_sound_effects

Sonorizá un video con Mirelo SFX 2.0, conservando el diálogo.

generate_music

Music with Suno V6, instrumental by default: two versions per request.

plan_film

Plans a product presentation or a motion-graphics reel: script, voice, music and scenes. Planning is free.

generate_film

Generates the voice, music, effects, images and clips of the film. From ~45 credits: the total is shown before anything is charged.

get_film

Film status and, when it is ready, the files and the steps to assemble the mp4 on your computer.

get_job

Status and result of a generation.

list_prompts

Library of tested prompts by category.

analyze_ads

Analysis of an ad library against twelve criteria.

score_creative

A 0 to 100 score for a creative before you run media.

scrape_meta_library

Brings in a brand's active ads on Meta.

How MCP works

A protocol so your assistant can use outside tools. Three steps and it is connected.

Choose your client

Follow the guide above and copy the URL or configuration for your client.

Authorize the connection

Ask your assistant

"With troquel, make a packshot of the cream in 4:5". Credits come off the same balance as the web.

Questions

What people ask us before connecting.

Get started

Does using MCP spend credits?+

Yes, exactly like the web. Same price per model, and if a generation fails the credits come back.

Where do I see what I made from my assistant?+

In your library. Everything that goes through MCP or the API is saved in your account with its prompt and model.

Do I need an API key?+

Claude, ChatGPT, Claude Code and Cursor can connect with your account. Hermes Agent, DeepSeek Harness and the API use a key. Follow the guide for your client.

How do I disconnect an assistant?+

In Settings, Connected apps, Disconnect. It loses access on its next call.

Can I have more than one key?+

Yes. One per client or per project, with a name, and you revoke them whenever you want.