PreMan field notes
Build for a new kind of software user.
Field notes on agent workflows, MCP tools, API fundamentals, and the infrastructure that keeps software reliable.
How PreMan works
Simulate. Probe. Repeat.
See how agents use your software, check workflows against your rules, and keep the tests that matter after each change.
- 01
Simulate
Recreate relevant conditions in a sandbox and define the rules your workflows should follow.
- 02
Probe
Probe MCP connections in parallel and exercise scenarios that check agent decisions.
- 03
Repeat
Turn failures into repeatable tests and rerun them as your tools and product change.
Start here
Product overview · 2 minWhat is PreMan? Infrastructure for the Agentic Era
PreMan helps keep software reliable as AI agents become its users, with parallel MCP probes, sandboxed workflow scenarios, and repeatable tests.
Inside the guide
- 01Software has a new kind of user
- 02Simulate relevant conditions
- 03Probe MCP connections in parallel
- 04Turn failures into repeatable tests
- 05Who PreMan is for
- 06Where we are going
The library
Build the rest of the system.
API fundamentals, testing playbooks, production reliability, MCP infrastructure, and practical comparisons.
From Git Push to Verified Fix: The Six-Stage Release-to-Repair Loop
A practical operating model for connecting a git push to API change detection, targeted tests, customer-impact analysis, a verified fix PR, and production monitoring.
What is an MCP? A Plain-English Guide to the Model Context Protocol
MCP (Model Context Protocol) is the open standard for letting AI models talk to your tools and data. Here is what it is, why it matters, and how to try one in 30 seconds.
What is an API? The 2026 Explainer
An API is a contract between two pieces of software. Here is what that means in plain English, what a real API call looks like, and the five types you will run into in 2026.
MCP vs API: When to Use Each
An MCP is an API designed for AI models to discover and call without your help. Here is a side-by-side comparison and a clear answer on when to use which.
How to Build Your First MCP Server in 10 Minutes
The shortest path from zero to a working MCP server that Claude can call. TypeScript SDK, working code, and the gotchas that will save you an hour.
How to Test an MCP Server, Step by Step
A seven-step process for testing an MCP server before you point a model at production. Handshake, schemas, auth boundaries, and the tests that only show up under a real model.
How to Build an API from Scratch (Node + Express)
A short, honest tutorial for building a working JSON API. Routing, validation, error handling, and what real production work looks like beyond the toy version.
Postman vs PreMan: Which Is Better in 2026
A practical comparison of Postman and PreMan in 2026. Where Postman still wins, where PreMan does, and how to decide.
The Best Postman Alternatives in 2026
A practical guide to Postman alternatives in 2026. What each tool is good at, what it gives up, and which is the right pick for your team.
Why Your MCP Server Isn't Connecting to Claude (and 9 Fixes)
If Claude Desktop shows your server but the tools do not appear, one of these nine causes is almost always the reason. With the fix for each.
How to Expose an Existing REST API to Claude or ChatGPT via MCP
You already have a REST API. You want Claude or ChatGPT to use it. Here is the cleanest path that does not require rewriting anything.
What is Tool Calling? Function Calls, MCP, and OpenAI Tools, Explained
Three terms that mean almost-but-not-quite the same thing. Here is how to keep them straight and which one to reach for.
Migrating from Postman to PreMan in Under 5 Minutes
A short walkthrough for moving a Postman workspace into PreMan. Export, import, refill secrets, run a request, done. Plus the few gotchas worth knowing.
How to Test AI Agents That Call Your APIs
An AI agent that calls your APIs is two products glued together. Both can fail. Here is the three-layer test plan that catches bugs at the seam.
Why MCP Is the New REST
A claim that gets thrown around too quickly. Here is the version of it that actually holds up, and the version that does not.
REST vs GraphQL vs MCP: A 5-Minute Breakdown
Three patterns, three jobs. Most teams will use at least two. Here is what each is good at and how to pick.
How to Write API Descriptions That LLMs Actually Understand
The single highest-leverage thing you can change to make AI agents pick the right tool. With before/after examples and a checklist you can ship today.
MCP Authentication: A Complete Guide
How auth actually works in MCP. The four real patterns (none, shared secret, per-user token, OAuth), what each is good for, and the security mistakes that keep showing up.
How to Audit an MCP Server: The Eight Checks That Matter
A practical, hands-on audit you can run against any MCP server in under an hour. The eight checks that catch most production bugs and the patterns that keep showing up across the ecosystem.
Is Postman Still Worth It in 2026? A Decision Framework
A practical framework for deciding whether to keep paying for Postman in 2026. Four kinds of teams that should stay, four that should leave, and the 15-minute audit that gives you the answer.
How to Maintain API Uptime When Every Minute Has a Price
99.99% uptime allows 4 minutes 19 seconds of downtime each month. Learn how to measure failures, protect the budget, and price an API SLA with evidence.
How to Write an Uptime SLA Guarantee That Survives an Outage
99.99% uptime leaves 4 minutes 19 seconds of monthly downtime. Write an API SLA with exact measurement, exclusions, claims, and credits that pay on time.
How to Account for Third Party Downtime in an SLA
Two 99.99% serial dependencies yield about 99.98% availability. Learn how to attribute vendor downtime and write fair SLA credit rules without hiding outages.
API Uptime Cost Explained: 95% to 99.999%
99.999% uptime permits only 25.9 seconds of monthly downtime. Compare a $1K, $4K, and $10K API SLA model with its real delivery costs before selling it.
How to Calculate API Uptime and Pay SLA Service Credits
12 failed minutes equal 99.9722% uptime in a 30-day month. Calculate exclusions, credit bands, and the exact SLA amount owed with a clear monthly receipt.
Five Agent Failures a Correct-Looking Answer Can Hide
A convincing answer can hide unauthorized actions, unchecked claims, stale state, account-boundary violations, and misread tool results. Here is what to test.
Build for agent users
Your product shouldn’t break when agents use it.
Simulate agents, probe MCP tools in parallel, and turn workflow failures into repeatable tests.
- 01Recreate relevant conditions
- 02Simulate agents in parallel
- 03Check workflow decisions
- 04Save repeatable tests