New MCP server support

Connect AI models together and ship the combo as an API

A simple, visual playground for building AI pipelines. Drag, connect, run — and pay only for the tokens you actually use.

No subscription · Pay as you go with crypto

Summarize and translate an article

The world's top models, ready to wire up on the canvas

OpenAI
Claude
Gemini
Llama
Anthropic
Perplexity
Hugging Face
NVIDIA
Ollama
ElevenLabs
Replicate

How it works

From idea to API in three steps

01

Drag and connect

Bring models, MCP tools and text templates onto the canvas and wire each output into the next input.

02

Try it right there

Run with one click and watch every node’s result, token usage and cost live on the canvas.

03

Publish and get an API

Every published version gets a stable endpoint. Call it from any language or service.

Feature tour

See every feature in a few seconds

Pick a feature to watch how it works, step by step.

Chat with models

1.Pick a model or a workflow

Pick any model or workflow and talk to it like a simple chat. Tokens and cost show up with every answer. Try it

Every model, one canvas

Claude, Gemini, GPT and custom models side by side — no juggling accounts and keys.

First-class MCP support

Paste an MCP server URL and its tools become nodes automatically.

Bring your own model

Upload your model’s Python code; it runs in an isolated sandbox and others can use it too.

Safe versioning

Editing a draft never touches the live API until you publish a new version.

Earn from what you build

Make a workflow public and take a fee on every call others make.

Transparent costs

Tokens and cost are recorded for every node on every run, so you know exactly where money goes.

Content automation

Hand your channels over to workflows

Build it once and the rest is automatic: the workflow runs on time, writes and creates the content with AI, and publishes it on YouTube, Instagram and Telegram.

  • Daily, weekly or every-few-minutes schedules, with a report of every run
  • Connect your accounts once; tokens are stored encrypted
  • Safe from runaway costs: after a few failures it stops itself and tells you

For developers

One workflow, one endpoint

Your canvas inputs and outputs become your API contract. Every response tells you exactly how many tokens were used and what it cost.

  • Separate, revocable API keys per service
  • Published versions are immutable — clients never break
  • Full run reports: input, output and cost per node
Read the docs
POST /v1/workflows/:slug/runbash
curl -X POST https://api.mlgrid.com/v1/workflows/article-digest/run \
  -H "Authorization: Bearer $MLGRID_API_KEY" \
  -d '{ "input": { "article": "..." } }'

{
  "status": "succeeded",
  "output": { "summary": "…", "english": "…" },
  "usage": { "tokensIn": 1840, "tokensOut": 380, "cost": "60.4", "currency": "IRT" }
}

Pricing

Pay only for the tokens you use

No subscription, no minimum. Top up your wallet and each run deducts exactly what each model consumed.

ModelInput / 1M tokensOutput / 1M tokens
Loading model list…
Cost calculator
200,000

Estimated cost

$0.00

Assuming 75% input and 25% output tokens

Frequently asked questions

Each model has a set price per million input and output tokens. After every run, the tokens actually used are multiplied by that model’s rate and deducted from your wallet. There is no monthly fee.

No. Building and running workflows is fully visual. You only need a few lines of code to call the API from your own software, and we provide ready-made samples.

MCP is an open standard for connecting models to tools and data. Add an MCP server and its tools (search, databases, external services…) appear as nodes on the canvas.

Every run happens in an isolated process with memory, CPU and time limits, with no access to platform keys or data.

Runs stop before exceeding your balance and the API returns 402. Top up with USDT/USDC and continue immediately.

Build your first workflow today

Sign-up takes less than a minute.

Start for free