Tensorify MCP Server for IDEs

The Tensorify Platform MCP Server gives AI agents full access to manage your Tensorify workspace — create workflows, deploy them, manage environment variables, and more. Use it from Cursor, Claude Desktop, or any MCP-compatible agent.

With the platform server, an AI agent can:

• Create and edit workflows using TSL (Tensorify Scripting Language)
• Deploy and undeploy workflows
• Execute workflows and check run results
• Manage environment variables
• Browse available plugins to understand what building blocks exist

Modern MCP clients (Cursor, Claude, and others) can connect without pasting an API key. Tensorify advertises OAuth 2.1 with PKCE (S256) and Dynamic Client Registration.

1. Open the app → Settings → MCP / Agents and copy the Cursor or Claude config (or use the JSON below).
2. Restart / reload the MCP client.
3. On first tool use, the client opens Tensorify's consent page — sign in, pick a workspace, approve scopes.

Cursor (.cursor/mcp.json or Cursor MCP settings):

{
  "mcpServers": {
    "tensorify-platform": {
      "url": "https://mcp.tensorify.io/v1/platform"
    },
    "tensorify-triggers": {
      "url": "https://mcp.tensorify.io/v1/triggers"
    }
  }
}

Claude Desktop / Claude Code (HTTP MCP):

{
  "mcpServers": {
    "tensorify": {
      "type": "http",
      "url": "https://mcp.tensorify.io/v1/platform"
    }
  }
}

Discovery endpoints (for custom clients):

• Resource metadata: https://mcp.tensorify.io/.well-known/oauth-protected-resource
• Authorization server: https://auth.tensorify.io/.well-known/oauth-authorization-server
• Consent UI: https://app.tensorify.io/oauth/authorize

Scopes: read:workflows, write:workflows, run:workflows.

For clients that cannot complete OAuth:

1. Go to Settings → API Keys and create a key with the scopes above.
2. Send Authorization: Bearer tfk_your_api_key on every MCP request.

{
  "mcpServers": {
    "tensorify": {
      "url": "https://mcp.tensorify.io/v1/platform",
      "headers": {
        "Authorization": "Bearer tfk_your_api_key"
      }
    }
  }
}

CLI helper: tensorify mcp config generates key-based snippets.

After connecting, the agent has access to these tools:

| Tool | Description | |------|-------------| | read_workflow_tsl | Get current workflow as TSL code | | apply_tsl | Create/replace workflow from TSL code | | patch_tsl | Incrementally modify a workflow | | validate_tsl | Check TSL code for errors |

| Tool | Description | |------|-------------| | list_projects | List all projects | | create_project | Create a new project | | list_workflows | List workflows in a project | | get_workflow | Get workflow metadata | | create_workflow | Create a new workflow | | delete_workflow | Delete a workflow |

| Tool | Description | |------|-------------| | deploy_workflow | Deploy a workflow | | undeploy_workflow | Stop a deployment | | execute_workflow | Trigger and get result | | get_run | Get run details | | list_runs | List recent runs |

| Tool | Description | |------|-------------| | set_env_var | Set an environment variable | | list_env_vars | List variable names | | delete_env_var | Remove a variable | | list_plugins | Available plugins | | get_plugin_manifest | Plugin details for TSL authoring |

TSL is the language agents use to create workflows:

import WebhookTrigger from "@tensorify/webhook-trigger"
import HTTPRequest from "@tensorify/http-request"
import Return from "@tensorify/return"

trigger = WebhookTrigger {
  path: "/my-api"
  method: "POST"
  responseMode: "use-workflow-response"
}

fetch_data = HTTPRequest {
  url: "https://api.example.com/data"
  method: "GET"
}

response = Return {
  body: "{{ fetch_data.response.body }}"
  status: 200
}

trigger.payload -> fetch_data.input
fetch_data.response -> response.input

• API keys are scoped — give agents only the permissions they need
• Keys can be revoked instantly from the dashboard
• All operations are scoped to your teamspace
• The MCP server validates permissions on every request

"Create a workflow that receives a webhook, fetches data from our API, transforms it, and sends a Slack notification"

The agent will:

1. Call list_plugins to see available plugins
2. Call get_plugin_manifest for the relevant plugins
3. Call create_workflow to make a new workflow
4. Call apply_tsl with the generated TSL code
5. Optionally deploy_workflow when ready

You can configure both the platform server and trigger server:

{
  "mcpServers": {
    "tensorify": {
      "url": "https://mcp.tensorify.io/v1/platform",
      "headers": { "Authorization": "Bearer tfk_..." }
    },
    "tensorify-tools": {
      "url": "https://mcp.tensorify.io/v1/triggers",
      "headers": { "Authorization": "Bearer tfk_..." }
    }
  }
}

This gives the agent both the ability to manage your workspace and call your deployed workflow tools.