Kong Context Management GA: Code Mode for agent MCP access

Kong announced Kong Context Management as generally available as part of Kong AI Management on 2026-09-30 (Alex Drag). The goal is letting agents reach more enterprise surface without stuffing every operation into the model context (Kong blog).

Code Mode: search → schema → execute

The core pattern is Code Mode. Instead of turning each API operation into its own MCP tool—Kong’s illustrative case is a 200-operation API becoming 200 tools—Code Mode exposes a compact MCP interface. The agent searches the available API surface, retrieves the relevant schema on demand, then executes. Breadth stays available; the full tool catalog does not have to sit in context on every turn (Kong blog).

Multi-step compose in code

Enterprise tasks often need more than one call. With Code Mode, Kong says agents can compose multiple operations in code—including loops, conditionals, and error handling—so fewer model↔tool round trips are required to finish a workflow (Kong blog).

What it covers

Context Management is framed as a way to create and manage agent access across APIs, existing MCP servers, and enterprise data sources, using Kong AI Gateway to generate MCP server interfaces. Kong’s cost / performance / quality triad is vendor product framing, not a measured independent bench (Kong blog).

Related posts and the URL path use Context Mesh as vision / stack branding. This write-up leads with Context Management—the GA product name on the announce—not Context Mesh as a second shipped SKU.

Pricing

The primary publishes no dollar figures, tiers, or SKUs. This brief invents none (Kong blog).

Who should care

Teams wiring agents to large enterprise APIs and MCP surfaces who want discover-then-execute access instead of per-operation tool bloat should start at the Kong Context Management post. Treat it as an AI Management / Code Mode GA, not a Kong AI Gateway feature dump and not a pricing card.