Model Context Protocol (MCP)

Your AI Client (Claude / Cursor / ...)
        │
        │  MCP Protocol (SSE)
        ▼
  Indite MCP Server   ──▶   Indite Platform
  mcp.indite.io                portal.indite.io
        │
        └── Tools: list_workspaces, create_workflow,
                   update_workflow, execute_workflow, ...

You:
I want a workflow that listens for incoming Telegram messages and replies with
an AI-generated response using Gemini. I have a Telegram bot credential saved
as "my-telegram-bot" and a Gemini credential as "gemini-prod" in my workspace.
Create this in my "Production" workspace, publish it, and activate it.

Claude:
Sure! Let me start by listing your workspaces to find the workspace ID...
[calls list_workspaces]

Found "Production" → ID: ws_abc123.
Now I'll create the workflow...
[calls create_workflow]
[calls get_workflow_block_schemas]
[calls update_workflow with telegram trigger → gemini block → telegram reply]
[calls publish_workflow]
[calls set_workflow_active]

✅ Done! Your Telegram bot is live. Send a message to your bot to try it.

You:
Run the "Fraud Detection" workflow in my "Analytics" workspace
with these values: order_id = "ORD-5021", amount = 3200, country = "NG".
Tell me the result.

Claude:
[calls list_workspaces → finds "Analytics" workspace]
[calls list_workflows → finds "Fraud Detection" → ID: wf_xyz789]
[calls execute_workflow with contextVariables]

Execution started — ID: exec_001. Polling for result...
[calls get_execution_status every few seconds]

✅ Result: FLAGGED
- Risk score: 0.91
- Reason: Unusual amount for this country/card combination
- Recommended action: Manual review

You:
Build a workflow that runs every day at 8 AM UTC. It should:
1. Call https://api.myapp.com/stats (GET, Authorization: Bearer {{apiKey}})
2. Send the JSON response to GPT-4o to write a human-friendly summary
3. Email the summary to reports@mycompany.com

Use my "openai-prod" credential. Workspace: "Operations".

Claude:
[calls list_workspaces → finds "Operations"]
[calls create_workflow → "Daily Stats Report"]
[calls get_workflow_block_schemas]
[calls update_workflow with:
  - schedule trigger (cron: 0 8 * * *)
  - httpRequest block → POST to stats API
  - openai block with GPT-4o to summarize
  - sendEmail block → reports@mycompany.com
]
[calls publish_workflow]

✅ Workflow created and published! It will run daily at 08:00 UTC.
I also added an {{apiKey}} variable — set its value in the workflow
variables panel if it isn't already configured.

You:
I need a workflow that:
1. Triggers via webhook with a refund request payload
2. Uses GPT-4o to evaluate whether the refund is valid (policy: orders < 30 days)
3. Pauses and asks for human approval if the amount > $500
4. If approved, calls https://api.payments.com/refund
5. Sends a Slack notification either way

Create it in "Customer Success" workspace.

Claude:
[creates workflow with webhook trigger → openai → condition (amount > 500) →
  human-input block → httpRequest → slack notification]
[publishes workflow]

✅ Done! The workflow is published. Here's the webhook URL to send refund requests to:
https://portal.indite.io/api/v1/webhook/wf_refund123

When a refund > $500 is received, the workflow will pause and notify your team.
Use this to resume it after review:

[calls get_execution_status to show how to check for PAUSED state]

To approve from Claude later, just say:
"Resume execution exec_abc with approved = true"

You:
Resume execution exec_refund_001 — approved = true, reviewed by "Sarah".

Claude:
[calls submit_human_input with blockId, output: { approved: true, reviewer: "Sarah" }]

✅ Execution resumed. The refund was processed and Slack was notified.