Skip to content

AGENTS.md Fresh

AGENTS.md (also available as AGENT.md) provides explicit guidance for AI agents, coding assistants, and LLM-based tools interacting with the X Developer Platform documentation. It is the recommended starting point for any agent.

Preferred documentation access methods

Always prefer these methods for the most accurate, up-to-date, and agent-friendly content:

  1. llms.txt (recommended starting point) - a small, curated root index linking to section-specific indexes:
    • x-api/llms.txt (X API v2 - 370+ pages)
    • enterprise-api/llms.txt
    • x-ads-api/llms.txt
    • xdks/llms.txt (Python + TypeScript XDKs)
  2. llms-full.txt - complete documentation as a single Markdown file for maximum context.
  3. Raw Markdown for any page - append .md to any documentation URL. Every page supports this; use it instead of the HTML view when possible.
  4. MCP Server - the full MCP server exposing 200+ X API endpoints plus documentation search.
  5. skill.md - a structured description of every action an agent can perform (agentskills.io format).

Site structure and navigation

  • X API v2 (/x-api/...): Posts, Users, Direct Messages, Lists, Spaces, Media, Streams (filtered + volume), Compliance, Webhooks, Account Activity, Trends, News, Usage, Connections.
  • Enterprise APIs (/enterprise-api/...): Account Activity (webhooks), X Activity (XAA), GNIP/PowerTrack historical and real-time, Compliance.
  • Ads API (/x-ads-api/...): Campaign Management, Creatives, Audiences, Analytics, Measurement, Catalog.
  • SDKs (XDKs) (/xdks/python/... and /xdks/typescript/...): official client libraries with full type coverage, pagination, and streaming helpers.
  • Fundamentals: Authentication, rate limits, data dictionary, expansions, fields, pagination, versioning, consistency.
  • AI & Agent Tools (/tools/ai, /tools/llms-txt, /tools/skill-md, /tools/mcp).

Important technical notes for agents

  • Authentication: OAuth 1.0a (user context), OAuth 2.0 (user context with PKCE, plus app-only Bearer), and Basic Auth for some enterprise endpoints.
  • Rate Limits: most endpoints have both app-level and user-level limits.
  • Data Model: use the official data dictionary, fields, expansions, and metrics. Posts, Users, and Spaces are the core objects.
  • Real-time Data: Filtered Stream, Volume Streams, Account Activity webhooks, and X Activity (XAA) are the primary real-time mechanisms.
  • Compliance & Safety: always respect developer terms, display requirements, and restricted use cases.

How to use the documentation effectively

flowchart TD
    A[Start: read AGENTS.md] --> B[Discover pages via llms.txt]
    B --> C[Fetch pages via .md suffix]
    C --> D{Need deep context?}
    D -->|Yes| E[Load llms-full.txt]
    D -->|Actions?| F[Load skill.md]
    E --> G[Use MCP for live calls]
    F --> G
    G --> H[Write code with the official XDKs]

Do not

  • Rely solely on the HTML-rendered pages when a clean Markdown alternative exists.
  • Assume deprecated v1.1 endpoints are still primary (focus on v2 and Enterprise equivalents).
  • Ignore rate limits, authentication context (app-only vs user context), or pagination requirements.
  • Generate code that violates the X Developer Agreement or Display Requirements.

See also