# Changelog of X (Twitter) Public Intel - MCP Server (`seibs.co/mcp-x-twitter-intel`) Actor

- **URL**: https://apify.com/seibs.co/mcp-x-twitter-intel/changelog.md
- **Full Actor documentation**: https://apify.com/seibs.co/mcp-x-twitter-intel.md

## Changelog

### 0.1 - 2026-06-12 - Initial release

- MCP server wrapper for `seibs.co/x-twitter-intel`.
- Four typed tools exposed to AI agents: `get_profile_intel`, `get_recent_posts`, `get_posts`, `compare_profiles`.
- Three run modes: `list_tools` (free tool catalog + agentic-payment descriptor), `call_tool` (one tool), `batch` (up to 10 `{tool, args}` calls).
- Slim projections over the real upstream dataset fields (`profile_record`, `post_record`, `engagement_velocity_signal`); newest-first ordering on recent posts; engagement-per-follower ranking on compare.
- Pydantic-validated args: handle cardinality (compare\_profiles 2-5), `max_posts` capped at 20, `post_inputs` 1-50.
- Pricing: flat $0.005 `mcp_tool_call` per invocation, plus upstream PPE pass-through billed to the same run.
- Cost-control: `_RunBudget` guard (600s compute cap, low-revenue abuse abort, hard cap of 5 tool calls), pre-flight input caps, batch cap of 10, hardcoded PPE prices in sync with `pay_per_event.json`.
- Agentic payments: x402 (USDC on Base) and Skyfire ready via `payments.py`; advertised in the `list_tools` catalog.
- Honest scope: about the 20 most recent posts per profile that X's public, logged-out syndication surface exposes - an intel tool, not a firehose. Logged-out public data only; PII minimized; not affiliated with X.