AI coding agents drive 66% of docs traffic
Mintlify released its midyear 2026 documentation traffic report, showing AI agent activity surged to 66% of documentation web traffic in July with over 213 million requests logged. An internal benchmark across 20 documentation sites revealed that providing an llms.txt file reduced agent error rates by nearly 90%.
AI coding agents are rapidly becoming the primary consumers of developer documentation, forcing a shift from human-oriented web pages toward structured, machine-readable routes like `llms.txt` and MCP.
* Agent web requests reached 66% of total documentation traffic in July 2026, up from 15.2% in January, and are projected to reach 90% by year-end.
* Explicit machine-readable routes account for 83.7% of agent traffic, with direct `.md` file requests growing to 54.4% of volume.
* Displaying an `llms.txt` file reduced agent retrieval error rates by nearly 90%, preventing wasted tool calls, dead URL hits, and excess token consumption.
* Hosted MCP servers processed 1.42 million tool calls across docs sites in July, with search tool calls accounting for 56.57% and direct filesystem reads taking 43.30%.
DISCOVERED
1h ago
2026-07-29
PUBLISHED
1h ago
2026-07-29
RELEVANCE
AUTHOR
mintlify