AI agent MCP handoff for tweet search & exports

Use this page when an AI agent needs to choose the right Xquik interface. It can search tweets, inspect profiles, export followers, or recover missed webhooks. Keep live calls narrow. Aggregate high-volume pages inside the sandbox. Store cursor state. Move durable jobs to REST, SDKs, or webhooks.

Pick the agent interface

Docs MCP

Use https://docs.xquik.com/mcp when the agent needs public docs, API reference pages, examples, troubleshooting, or type definitions. It is read-only and requires no auth.

API MCP

Use API MCP at https://xquik.com/mcp for live calls. Full account keys and OAuth tokens expose every catalog route. Guest paid_reads keys expose only the eligible GET routes. Current SDKs negotiate MCP 2026-07-28 through server/discover.

REST or SDK

Use REST or generated SDKs when a service owns retries, cursor storage, file downloads, queues, or batch jobs outside the chat session.

Webhooks and replay

Use monitor webhooks for fresh events and GET /api/v1/events when a receiver, queue, warehouse, or agent run needs replay.

Use the default agent route

Search public docs or llms.txt before requesting current API data.
Use docs for product guidance. Use search before xquik.request(...) to confirm the endpoint path, required parameters, costs, and response shape.
Keep requested filters. Never add others. Call xquik.request({ path, method?, query?, body? }) with the smallest useful page. Hosted MCP injects authentication and required idempotency headers.
Continue while has_more is true and next_cursor advances.
Return normalized rows, IDs, counts, samples, and cursors instead of full raw pages.
Hand long-running or replayable work to REST, SDKs, webhooks, or exports.

Credential lifecycle operations and direct saved-payment mutations are not in the MCP catalog. Guest wallet creation, status, and top-up also remain direct REST only. A 402 creates no checkout. Report its payment choices, ask the user to choose an amount and option, then wait for explicit confirmation. A full account session may execute only an advertised account checkout action in its catalog. Never execute guest wallet routes through MCP.

async () => {
  return Object.entries(spec.paths).flatMap(([path, methods]) =>
    Object.entries(methods).flatMap(([method, operation]) =>
      path.includes("/x/tweets/search") || operation?.summary?.toLowerCase().includes("followers")
        ? [{ method: method.toUpperCase(), path, ...operation }]
        : [],
    ),
  );
};

Route guest paid reads

An active guest key gives search and execute a read-only catalog of the eligible operations. The docs tool remains available. Every route appears in the guest paid-read inventory. Batch GET /api/v1/x/tweets accepts up to 100 tweet IDs. The sandbox cannot execute writes, account actions, automations, billing, credential management, or any other read. OAuth and full account behavior remain unchanged. If the guest key needs creation, activation status, or more credits, leave MCP and follow the accountless guest wallet flow through direct REST:

Creation and top-up require explicit user confirmation. Status polling does not create payment. Never place the guest key or its creation Idempotency-Key in tool code, chat output, or shared agent state.

Return stored rows

MCP returns normalized snake_case fields, structured errors, and date-time fields as Unix seconds. Keep agents on has_more and next_cursor even when REST or SDK pages show camelCase response fields. Read a REST createdAt field as created in MCP results.

async () => {
  const query = "from:username MCP";
  const { result: page } = await xquik.request({
    path: "/api/v1/x/tweets/search",
    query: { q: query, limit: "50" },
  });

  return {
    source: "xquik_mcp",
    job: "tweet_search",
    query,
    rows: page.tweets.map((tweet) => ({
      tweet_id: tweet.id,
      text: tweet.text ?? null,
      author_id: tweet.author?.id ?? null,
      author_username: tweet.author?.username ?? null,
      created_at: tweet["created"] ?? null,
      url: tweet.url ?? null,
    })),
    has_more: page.has_more,
    next_cursor: page.next_cursor,
  };
};

Follow cursor rules

Source	Next request
MCP tweet, profile, follower, reply, timeline, community, and list pages	Pass `next_cursor` back as `cursor` when `has_more` is true.
MCP `/api/v1/draws`, `/api/v1/extractions`, and `/api/v1/events`	Pass `next_cursor` back as `cursor`.
MCP `/api/v1/radar`	Pass `next_cursor` back as `after`.
MCP `/api/v1/drafts`	Pass `next_cursor` back as `afterCursor`.
REST and generated SDKs	Follow the response fields documented on that endpoint page.

Full account sessions can check GET /api/v1/credits before large reads. Guest sessions receive balance and top-up context in 402. The direct REST guest status route reports the balance. Low balances can return smaller pages. Zero affordable rows can return 402 insufficient_credits. For high-volume MCP reads:

Deduplicate tweet or user rows by stable id
Continue through empty pages when has_more is true
Stop at the requested total, page cap, has_more: false, or cursor_stalled when cursors repeat. Keep output below 24,000 characters.

Store handoff state

Store the values a later agent, service, or workflow needs to resume without reading chat history.

{
  "agent_job_id": "mcp-research-q2",
  "surface": "api_mcp",
  "endpoint": "GET /api/v1/x/tweets/search",
  "query": "from:username MCP",
  "cursor_param": "cursor",
  "next_cursor": "DAACCgACGRElMJcAAA",
  "has_more": true,
  "webhook_id": "15",
  "delivery_id": "502",
  "stream_event_id": "9002",
  "event_replay_route": "GET /api/v1/events?cursor=9002",
  "export_route": "GET /api/v1/extractions/77777/export?format=json",
  "saved_fields": ["tweet_id", "author_username", "text", "url"]
}

Keep API keys, webhook secrets, raw request bodies, raw signatures, and full headers out of chat transcripts, shared agent memory, spreadsheets, CRM rows, and queue payloads.

Know when to leave MCP

File exports

Use extraction export endpoints for CSV, JSON, XLSX, Markdown, or PDF files.

Webhook receivers

Use signed webhooks when downstream systems need fresh monitor events.

Replay jobs

Use stored events and delivery rows when receivers miss work.

SDK backends

Use SDKs when a backend owns retries, storage, and batch orchestration.

Continue with focused references

MCP tools reference

Review Code Mode, OpenAPI-native tools, response contracts, and examples.

Docs MCP server

Connect read-only documentation search beside the API MCP server.

No-code handoff

Hand monitor events, exports, and direct reads to workflow platforms.

Webhook testing

Verify signed receivers before accepting production events.

Getting started

SDKs

Guides

Webhooks

MCP server

Agent plugins

Framework guides

OAuth

MPP

Resources

AI agent MCP handoff for tweet search & exports

Pick the agent interface

Docs MCP

API MCP

REST or SDK

Webhooks and replay

Use the default agent route

Route guest paid reads

Return stored rows

Follow cursor rules

Store handoff state

Know when to leave MCP

File exports

Webhook receivers

Replay jobs

SDK backends

Continue with focused references

MCP tools reference

Docs MCP server

No-code handoff

Webhook testing

​Pick the agent interface

Docs MCP

API MCP

REST or SDK

Webhooks and replay

​Use the default agent route

​Route guest paid reads

​Return stored rows

​Follow cursor rules

​Store handoff state

​Know when to leave MCP

File exports

Webhook receivers

Replay jobs

SDK backends

​Continue with focused references

MCP tools reference

Docs MCP server

No-code handoff

Webhook testing

Pick the agent interface

Use the default agent route

Route guest paid reads

Return stored rows

Follow cursor rules

Store handoff state

Know when to leave MCP

Continue with focused references