MCPbundler

Blog

Insights on MCP servers, AI agent integration, and building with Model Context Protocol

Engineering
mcptool-routingscaling

What Actually Happens When One MCP Server Has 200 Tools

An honest audit of what MCP Bundler solves for tool-heavy servers today, and what's genuinely still a gap - going on the roadmap, not glossed over.

MCP TeamAugust 10, 2026 - 4 min read
Engineering
updatesi18n

How we keep two languages honest without a cron job

Tracking translation staleness with a content hash instead of a timestamp, and why the pipeline still refuses to publish its own translations.

MCP TeamAugust 9, 2026 - 2 min read
Engineering
mcpprotocolspec

Is MCP Stateless Now? What the 2026-07-28 Spec Actually Changed

MCP dropped its session handshake in the 2026-07-28 spec. Here's what actually changed, why 'stateless' undersells it, and the new trust boundary nobody's leading with.

MCP TeamAugust 9, 2026 - 6 min read
Engineering
mcpcontext-windowperformance

One MCP Call, 24K Characters: Why Tool Results Keep Blowing Up Your Context Window

A single MCP tool call putting 24,568 characters into someone's context isn't a fluke - it's the default outcome of how most MCP servers respond. Here's why, and what actually fixes it.

MCP TeamAugust 9, 2026 - 6 min read
Engineering
updatesseoai-agents

We shipped llms.txt - here's exactly what's in it, and what isn't

Most llms.txt files are a markdown file nobody validates. Ours is generated from live content - and we're telling you the one thing it doesn't cover yet.

MCP TeamAugust 9, 2026 - 2 min read
Engineering
updateswebhooksintegrations

Webhooks that fail loud, not quiet

Most webhook systems are honest about signing and vague about everything else. Ours tells you exactly when it'll retry, when it won't, and why.

MCP TeamAugust 9, 2026 - 3 min read