Kong
Second with 3 out of 5. An open-source core with plugins, popular with teams running Kubernetes.
Microsoft's managed gateway enforced every usage limit we set and barely slowed traffic, but it ties you to Azure and the larger tiers are pricey.
Of the seven gateways we put through the same test, Azure API Management was the one we'd trust with real customers. It finished first in our API management comparison with 4.5 out of 5, and it led because of what it got right rather than any single flashy feature.
Between October 3 and October 16, 2026, a 14-day run, it passed 27 of our 30 gateway checks and enforced all 15 usage limits we wrote. It added about 25 milliseconds to the slowest requests, and we could export our whole set-up without rebuilding anything by hand.
The fine print: it only runs on Microsoft's cloud, its rules are written in XML, and the production tiers start around $147 a month and climb to almost $2,800. If your company already lives on Azure, it's the obvious starting point. If not, weigh the lock-in first.
Features earned a perfect 5. Developer experience was the softest area at 3.5, mostly because getting started still took over an hour.
Overall: 4.5 out of 5, the mean of the five parts rounded to the nearest half point.
See the six tools Azure beat, and which one might still suit you better.
Picture the reception desk of a busy office building. Visitors don't wander straight to the desk they want; they sign in at reception, show a badge, get pointed to the right floor, and reception keeps a log of who came and went. If a crowd turns up at once, reception slows the queue so the people upstairs aren't swamped.
An API gateway is that reception desk for software. An API is the way one program asks another for data, for example a shopping app asking your server for stock levels. The gateway sits in front of your servers and checks each request's credentials (authentication), caps how many requests one customer can make (rate limiting), sends each request to the right place and records it all for reports (analytics).
Azure API Management is Microsoft's hosted version. It comes in three pieces: the gateway itself, an admin area in the Azure portal where you write the rules, and a developer portal, a ready-made website where outside developers sign up, read your docs and try calls. It sits among the more technical products in our developer tools section, so expect to need someone comfortable with cloud set-up.
Everything happens in the Azure portal, Microsoft's web console. Here's the route we took.
1. Create an instance. An instance is your own copy of the service. You choose a tier (see pricing below), a region and a name. Classic tiers can take a while to create, so start this first.
2. Bring in your APIs. Rather than describing each API by hand, you import a description file. Azure accepts the common formats, such as OpenAPI (the most widespread), WSDL, GraphQL and gRPC.
3. Write policies. Policies are the rules: require a key, allow 100 calls a minute, rewrite a header. They're written as short XML snippets, a structured text format that looks fussy at first but follows a pattern.
4. Package and share. You group APIs into products, for example a free product and a paid one with higher limits, and publish them on the developer portal.
Not every feature is on every tier, which is the main thing to check before choosing one.
Authentication, rate limits, caching, request rewriting and logging. Every one of our 15 rate-limit rules behaved as written.
A customisable website with docs and a "try it" console. Included on every classic tier except Consumption.
Runs the gateway as a container on your own servers or another cloud. Free on Developer (one copy per deployment); an add-on on Premium. Not offered on the v2 tiers.
Let separate teams run their own APIs inside one instance. A Premium feature, and each workspace needs its own workspace gateway.
Premium can run in several Azure regions and availability zones, so one data-centre problem doesn't take your APIs down.
Controls traffic to AI models and MCP servers, the connectors AI agents use. Microsoft made this generally available in November 2025.
Our Azure API Management test covered two weeks, October 3 to October 16, 2026. Each gateway in this group protects the same three sample APIs, and we measure the same things on all of them: time to get running, how well usage limits hold, what the reports show and how much delay the gateway adds. Our testing method explains the set-up in full.
Getting started took 67 minutes using only Microsoft's documentation. That's not instant, but nothing went badly wrong: the docs passed 9 of our 10 quality checks and 4 of the 5 sample set-ups worked as printed. We had to fix the fifth ourselves.
Then came the main job. Across the three APIs it passed 27 of 30 gateway checks, the strongest result in the group. We wrote 15 usage-limit rules, the kind that stop one customer from hammering your servers or keep free users on a lower allowance than paying ones. All 15 behaved exactly as written. Several rivals let a third or more of these rules slip.
Speed was the other highlight. We looked at the slowest 5% of requests (what engineers call the 95th percentile), and the gateway added only about 25 milliseconds to them. For comparison, a blink takes roughly ten times that. Last came the exit test, which asks whether you could move to another provider later: all 5 export checks passed and we needed no manual rebuilding.
| Measure | Result | In plain terms |
|---|---|---|
| Time to first working request, docs only | 67 min | About an hour of set-up before anything works. |
| Gateway checks passed, three APIs | 27/30 | Nine in ten behaved as configured, the best of the group. |
| Docs and sample set-ups | 9/10 doc checks; 4/5 samples work | Thorough docs, with one sample that needs fixing. |
| Delay added, slowest 5% of requests | ≈25 ms | Your users won't feel it. |
| Export and leaving | 5/5 checks; 0 manual migrations | You can move elsewhere without rebuilding by hand. |
None of the weak points we found came from the test itself. They're about cost once you outgrow the cheap tiers, and about committing to Microsoft's cloud.
Tested October 3, 2026 to October 16, 2026.
Every one of our 15 usage-limit rules held. If those limits protect your servers or separate free and paid customers, this matters more than any feature list.
About 25 ms added to the slowest requests, alongside 27 of 30 checks passed. The gateway does its job without getting in the way.
All 5 export checks passed with nothing to rebuild by hand, so you're less locked in than the Azure-only hosting suggests.
Developer is about $48 a month, but production use starts at about $147 on Basic and reaches about $2,795 on Premium, per unit.
You can't move the managed service to another cloud, and writing policies in XML puts off developers used to simpler formats.
67 minutes to a first request and one broken sample set-up. Fine for a team project, slow if you hoped for a quick trial.
Microsoft sells two families of tiers. The classic tiers include Consumption, which charges per call, and fixed monthly tiers billed per unit, a unit being one slice of capacity. The newer v2 tiers bundle a monthly allowance of requests. Prices below are the US figures on Microsoft's pricing page and vary by region.
Consumption is free for the first 1 million API operations per subscription each month, then $0.042 per 10,000 operations, or $4.20 per million. Among the classic fixed tiers, Developer costs $48.04 a month, Basic $147.17, Standard $686.72 and Premium $2,795.17, each per unit. In the v2 family, Basic v2 is $150.01 a month with 10 million requests, Standard v2 $700 with 50 million, and Premium v2 $2,801 with unlimited requests.
Premium adds multiple regions, availability zones and workspaces, and it can reach a 99.99% uptime promise when spread across two or more zones or regions. Running the self-hosted gateway on Premium costs an extra $250.01 a month per deployment. An Isolated tier is in preview, priced through Microsoft support.
Developer, Basic, Standard and Premium bill by the hour while an instance exists, even if nobody calls your APIs, so delete test instances when you're done. We didn't find refund terms specific to API Management.
Prices checked on November 9, 2026 on Microsoft's official pricing page. They may have changed since.
Every other gateway scored lower. Kong came second with 3 out of 5; it grew out of a free open-source gateway, with paid plans from $25 a month, and suits teams running Kubernetes. MuleSoft also scored 3 and mixes API control with connecting business apps, but starts around $2,000 a month.
The nearest like-for-like rival is Apigee, Google Cloud's gateway, which scored 2.5. It charges about $365 a month per region before any calls, while Azure's Consumption tier lets small projects start for nothing. Our gateway ranking puts all seven side by side.
| Tool | Starting price | Who it's for |
|---|---|---|
| Azure API Management | Free (Consumption); Developer $48.04 a month | Teams on Microsoft Azure |
| Kong | $25 a month | Platform teams on Kubernetes or hybrid set-ups |
| MuleSoft | From $2,000 a month | Large firms combining app integration and API control |
| Apigee | About $365 a month | Large organisations on Google Cloud |
Because it's a Microsoft service, it plugs into Microsoft's own security tools: Entra ID for sign-in, Key Vault for storing secrets and Defender for APIs for threat alerts. Uptime promises (called SLAs) are 99.95% on Consumption, Basic and Standard, and up to 99.99% on Premium when spread across zones or regions. We couldn't confirm certifications or data-handling details from the pages we read.
Two practical warnings. The Developer tier has no uptime promise at all, so never put paying customers on it. And Microsoft retires features on a schedule: trusted service connectivity ended on March 15, 2026 and isn't available for services created after December 1, 2025, and the AI gateway preview began moving to a new resource type on October 14, 2026.
Check Microsoft's retirement announcements before you build on a preview feature.
Feedback is positive but cooler than our test. On Gartner Peer Insights, a site for verified IT buyers, it showed 4.1 out of 5 across 230 reviews at our November 3, 2026 reading. On G2 it had 4.2 out of 5 from 30 reviews the same day. We couldn't recheck either site on November 9, so both figures date from November 3.
Reviewers like the sign-in, rate-limiting and security rules, and how smoothly it works with other Microsoft services. The complaints: cost for smaller teams, a learning curve for advanced policies and the developer portal, limits on custom code inside policies, and occasional network errors that caused downtime.
The network errors some reviewers mention didn't appear in our 14 days, but two weeks is short, so monitor your gateway closely in its first months of real traffic.
These are the gateways that came closest, with their scores from the same test.
Second with 3 out of 5. An open-source core with plugins, popular with teams running Kubernetes.
3 out of 5. Joins up business apps and governs APIs in one platform, often alongside Salesforce.
3 out of 5. Not really a gateway: it gives developers a single API for email, calendars, contacts and tasks.
2.5 out of 5. Google Cloud's counterpart, aimed at large organisations already on Google.
2.5 out of 5. A free open-source gateway for technical teams happy to host it themselves.
If your systems already run on Azure, this is the gateway to try first. In our test it passed 27 of 30 checks, enforced all 15 usage limits, added only about 25 ms of delay and let us leave cleanly. Those results earned 4.5 out of 5 and first place among seven gateways.
Its drawbacks sit outside the test: you're committing to Microsoft's cloud, you'll write rules in XML, and the production tiers get expensive quickly. Start on Consumption, or Developer if you need the portal, and move up only when real traffic demands it. Teams that want to stay cloud-neutral should look at Kong.
Azure API Management suits Azure-based teams that need dependable limits and low delay; anyone avoiding Microsoft lock-in should look elsewhere.
All seven tools protected the same three APIs. The full ranking compares checks, delay and price.
If you use Azure API Management, or work there, and think we got something wrong, email us. Tell us what you saw and attach screenshots if you have them. We'll take another look and update the review if it changes our findings.
Partly. The Consumption tier includes the first 1 million API operations per Azure subscription each month at no charge, then costs $0.042 per 10,000 operations. It has no developer portal, though. The other tiers have a fixed monthly price.
On Microsoft's pricing page on November 9, 2026, the classic tiers cost $48.04 a month for Developer, $147.17 for Basic, $686.72 for Standard and $2,795.17 for Premium, per unit. The v2 tiers cost $150.01 (Basic v2), $700 (Standard v2) and $2,801 (Premium v2) a month. Prices vary by region.
The v2 tiers bundle a monthly request allowance: 10 million on Basic v2, 50 million on Standard v2 and unlimited on Premium v2, with a fee per extra million on the first two. They don't offer the self-hosted gateway, which is only on the classic Developer and Premium tiers.
Partly. The managed service lives in Azure, but the self-hosted gateway lets you run the traffic-handling part as a container on your own servers or another cloud. It's free on the Developer tier and costs $250.01 a month per deployment on Premium.
It isn't designed for that. Microsoft gives the Developer tier no uptime promise and describes it as for non-production use and evaluation. For real customers, use Basic or higher, which promise 99.95% uptime.
Our two-week test started on October 3, 2026 and ended on October 16, 2026. In that time we put three sample APIs behind it using Microsoft's documentation, timed the set-up, ran 30 gateway checks and 15 usage-limit rules, measured the added delay and checked how easily we could export our set-up.