Knowledge
Our Thoughts
What is PHP2Shell, and is your website at risk?
What the PHP2Shell flaws are, why so many PHP sites are exposed, and why a Next.js site on Vercel does not run the code they affect.
How long can a Vercel Function run, and can it hold a WebSocket?
Vercel Function timeouts on Fluid compute, why an 800-second maxDuration can fail a deploy, and how we run WebSocket live cursors on Vercel Functions.
Is Vercel AI Gateway worth it for production AI apps?
How Vercel AI Gateway handles fallbacks, BYOK and per-customer spend, how it compares with OpenRouter, and where it falls short, from Monogram's production use.
Is Vercel eve ready for production agents?
What Vercel eve is, how it differs from the AI SDK and LangGraph, whether it needs Vercel, and where it stops, from the eve agents Monogram runs in production.
What is Vercel BotID, and when should you use it instead of a CAPTCHA?
What Vercel BotID does, how it compares with Turnstile and reCAPTCHA, Basic vs Deep Analysis, and how we run it on a public share page in Monogram Tools.
When does a background job need Vercel Workflows instead of a single function?
When to move a job from after() or a cron onto Vercel Workflows, how steps, retries and hooks behave, and the one long job we kept off it on purpose.
How does Vercel Connect keep OAuth tokens out of your environment variables?
What Vercel Connect is, how getToken, installations and token refresh work, how it compares with Nango or storing tokens yourself, and how we run Slack on it.
Where should files live in a Vercel app, and how much work should a cron do?
How we keep files in private Vercel Blob, and when a Vercel Cron Job should do bounded work itself or reconcile and hand off to a durable run.
From 72 to 99 on Is Agentic: what we changed on monogram.io
On Aug 27, Vercel's Is Agentic scanner scored monogram.io 72 out of 100. 11 days and one pull request later it reads 99. What the scanner measures, what we changed, and the checks that still fail.
SEO vs. AEO: What AI visibility requires now
Search now produces direct answers, summaries, comparisons, citations, and recommendations. AEO focuses on whether your information can be found, understood, trusted, and reused inside those experiences.
Can Vercel OIDC federation replace the keys in your environment variables?
How Vercel OIDC federation issues short-lived tokens for AWS, GCP, Azure and Vercel's own services, why local tokens expire, and the gotchas we hit using it.
When should code run in Vercel Sandbox instead of a Vercel Function?
When to move work from a Vercel Function into Vercel Sandbox, how snapshots remove cold installs, and the limits we hit running microVMs in production.
Should you use the AI SDK or LangChain in TypeScript?
When to pick the AI SDK over LangChain or a provider SDK, how it handles RAG and structured output, and what the v7 upgrade costs, from Monogram's code.
We ship the AI layer with the code
When we hand a repo back, the AI layer we built with comes with it: instruction files, specs, skills, a DESIGN.md, and an MCP server. What each one does, and why the client's team can keep building without us.
Spec-Driven Development: Turning AI Coding from Guesswork into Ground Truth
An agent can build the feature and still have no idea why it exists.
Meritech Capital
Portfolio Reel
What isitagentready.com Actually Measures
Kiro: The (Power)ful Agentic IDE
From Vector Search to Knowledge Runtime: The New RAG Standard
Declarative macOS with Nix: A Practical Guide