Skip to main content
Most no-code platforms do not need to run the MCP session themselves. A reliable pattern is to let an MCP-capable agent collect data, then hand structured JSON to the workflow.

​
Prerequisites

  • A Twexapi API key
  • An MCP-capable agent or framework connected to https://api.twexapi.io/mcp
  • A webhook URL from n8n, Zapier, Make, Pipedream, or your internal workflow runner
  • A destination system such as Sheets, Airtable, Notion, Slack, a CRM, or a database
1

Collect with an MCP-capable agent

Use Claude Code, Cursor, Codex CLI, LangChain, Pydantic AI, or another MCP client connected to https://api.twexapi.io/mcp.
2

Return strict JSON

Ask the agent for a stable schema with IDs, cursors, route names, and rows.
3

Send to the no-code tool

Post the JSON to a webhook in n8n, Zapier, Make, Pipedream, or your internal workflow runner.
4

Continue through REST if needed

Scheduled or high-volume jobs can call Twexapi REST directly with Authorization: Bearer YOUR_API_KEY.

​
Copy-paste agent prompt

Use Twexapi MCP to search recent X/Twitter posts about "AI agents".
Call explore first, then twexapi_request.
Return only valid JSON with:
{
  "route_used": "...",
  "query": "AI agents",
  "has_more": true,
  "next_cursor": "...",
  "tweets": [
    {
      "tweet_id": "...",
      "author_username": "...",
      "created_at": "...",
      "text": "...",
      "like_count": 0,
      "reply_count": 0,
      "retweet_count": 0
    }
  ]
}
Do not call write endpoints.

​
Workflow mapping

JSON fieldCommon destination
tweet_idDatabase primary key, dedupe key, CRM note reference
author_usernameCRM account field, Slack message label
created_atSort key and reporting timestamp
next_cursorWorkflow state for the next scheduled run
route_usedAudit log and debugging field

​
Result handoff

Keep the workflow payload small and stable. Do not pass the full raw response to Slack messages, CRM notes, or spreadsheet rows unless you also archive it separately.
Row typeRequired fields
Tweet searchtweet_id, text, author_username, created_at, public_url, route_used
Profile lookupuser_id, username, name, description, followers_count, verified
Trend rowsname, rank, query, requested country, requested topic
Paginationhas_more, next_cursor, original query or source ID
Write planaction, endpoint, preview, requires_human_confirmation

​
Recipe: Agent research to webhook

  1. Connect an AI agent to Twexapi MCP.
  2. Ask it for strict JSON using the prompt above.
  3. POST the JSON to your workflow webhook.
  4. Use the workflow tool’s iterator or loop step over tweets.
  5. Upsert rows by tweet_id.

​
Recipe: Scheduled continuation

For daily jobs, keep the first run agent-assisted and make later runs deterministic:
  1. Store route_used, request parameters, and next_cursor.
  2. Run scheduled calls through Twexapi REST with Authorization: Bearer YOUR_API_KEY.
  3. Stop when has_more is false or the workflow reaches its row limit.
  4. Send exceptions back to an agent only when the selected endpoint or query needs to change.

​
Testing checklist

  • Confirm the webhook receives valid JSON, not Markdown.
  • Confirm tweet_id or user_id is used for dedupe.
  • Confirm next_cursor is persisted outside chat history.
  • Confirm write actions are represented as plans until a human approves them.
  • Confirm API keys, cookies, and auth tokens are never sent to user-visible destinations.