Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions content/blog/inkan-0-10.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,6 @@ title: "Your API, as tools a model can call"
description: "inkan 0.10 serves every route as an MCP tool, made from the contract the route already has, and checks that what a model gets is what the docs say."
date: "2026-11-13"
author: "Vensin"
draft: true
---

Every inkan route carries its contract: what goes in, what comes out, and a few examples of
Expand Down Expand Up @@ -62,8 +61,10 @@ server as a process. `-H "authorization: Bearer …"` hands it a token. The docs

Without `mcp` nothing changes on the request path. While measuring I found that
`app.inject` had picked up an extra turn per request, and fixed it: the in-process bench is now
about 4 % faster than 0.9.4, almost all of it in `inject`, which only your tests use. The
numbers over the network are in the release notes.
about 4 % faster than 0.9.4, almost all of it in `inject`, which only your tests use. Over
the network, on the GitHub runner, inkan scores 89.8 against 0.9.4's 89.7 and Fastify's
88.6 (median of three runs), ahead of Fastify in all three but not in every scenario. The
[benchmarks page](/benchmarks) has the table.

## Next

Expand Down
19 changes: 10 additions & 9 deletions content/docs/performance.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -15,23 +15,24 @@ after a warm-up, 3 rounds each; the median of three such runs:
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| **inkan on uWebSockets.js** ¹ | **134.4** | 147 % | 148 % | 104 % | 99 % | 149 % | 154 % | 115 % |
| node:http, by hand | 100 | 100 % | 100 % | 100 % | 100 % | 100 % | 100 % | 100 % |
| **inkan** | **89.4** | **94 %** | 89 % | 86 % | 90 % | 89 % | **93 %** | 74 % |
| **inkan, sealed** | 88.3 | 85 % | **92 %** | **98 %** | 88 % | 89 % | 88 % | 76 % |
| Fastify | 87.2 | 88 % | 91 % | 91 % | 76 % | **90 %** | 87 % | 59 % |
| Hono | 82.5 | 84 % | 71 % | 94 % | **99 %** | 71 % | 74 % | **84 %** |
| Express | 46.1 | 44 % | 46 % | 71 % | 72 % | 23 % | 22 % | 50 % |
| **inkan** | **89.8** | 90 % | **93 %** | 87 % | 88 % | 91 % | **95 %** | 81 % |
| **inkan, sealed** ² | 88.3 | 85 % | 92 % | **98 %** | 88 % | 89 % | 88 % | 76 % |
| Fastify | 88.6 | 92 % | 90 % | 93 % | 80 % | **94 %** | 92 % | 76 % |
| Hono | 85.6 | **94 %** | 77 % | 93 % | **98 %** | 77 % | 85 % | **87 %** |
| Express | 48.9 | 53 % | 52 % | 69 % | 73 % | 23 % | 23 % | 53 % |

¹ With [`@inkanjs/uws`](/docs/integrations/uws), measured with 0.6.0 in a run of its own, against `node:http` in that run.
² Measured with 0.9.0, in the runs of that release.

The score is the geometric mean over all twelve scenarios against bare `node:http`,
so no single one can carry or sink a server. inkan does a little more per request
than the others: it gives every request an id and keeps back every key an answer's
contract does not list. Read the numbers as a range, not a rank: a run on another
day moves each server by a few points. inkan is a little ahead of Fastify: in front in all
three runs, by about two points, but not in every scenario (Fastify leads the body of 50, the
async handler and 3 middlewares). As fast as 0.8. With `inkan seal`, a body of 50 objects is
checked faster than by any of them. The runs of 0.9.0 are in
[`bench/reports/0.9.0.md`](https://github.com/inkanjs/inkan/blob/main/bench/reports/0.9.0.md).
three runs, by one to two points, but not in every scenario (Fastify leads the body of 50, the
400 routes and 3 middlewares). As fast as 0.9. With `inkan seal`, a body of 50 objects is
checked faster than by any of them. The runs of 0.10.0 are in
[`bench/reports/0.10.0.md`](https://github.com/inkanjs/inkan/blob/main/bench/reports/0.10.0.md).

Run it yourself: `cd bench && npm install && node run.mjs`, or the `bench` workflow
on GitHub, which also measures what one request costs the server in CPU time.
17 changes: 9 additions & 8 deletions data/bench.ts
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@
// A new run: change `run` and the values, nothing else on the page needs to move.

export const run = {
date: "2026-11-03",
date: "2026-11-13",
machine: "GitHub Actions runner, 4 cores",
tool: "autocannon",
seconds: 15,
Expand Down Expand Up @@ -35,14 +35,15 @@ export type Server = {
values: Record<ScenarioKey, number>;
};

// inkan 0.9.0: the median of three runs of the release bench, every server in each run;
// inkan on uWebSockets.js still from its own run with 0.6.0, against node:http in that run.
// inkan 0.10.0: the median of three runs of the release bench, every server in each run;
// inkan sealed from the runs of 0.9.0, and inkan on uWebSockets.js from its own run with
// 0.6.0, each against node:http in its run.
export const servers: Server[] = [
{ name: "inkan on uWebSockets.js", what: "@inkanjs/uws, measured with 0.6.0", ours: true, score: 134.4, values: { hello: 147, params: 148, body: 104, answer: 99, router: 149, notFound: 154, invalid: 115 } },
{ name: "node:http, by hand", what: "the baseline", base: true, score: 100, values: { hello: 100, params: 100, body: 100, answer: 100, router: 100, notFound: 100, invalid: 100 } },
{ name: "inkan", what: "on node:http", ours: true, score: 89.4, values: { hello: 94, params: 89, body: 86, answer: 90, router: 89, notFound: 93, invalid: 74 } },
{ name: "inkan, sealed", what: "inkan seal", ours: true, score: 88.3, values: { hello: 85, params: 92, body: 98, answer: 88, router: 89, notFound: 88, invalid: 76 } },
{ name: "Fastify", what: "schemas, like inkan", score: 87.2, values: { hello: 88, params: 91, body: 91, answer: 76, router: 90, notFound: 87, invalid: 59 } },
{ name: "Hono", what: "checks by hand", score: 82.5, values: { hello: 84, params: 71, body: 94, answer: 99, router: 71, notFound: 74, invalid: 84 } },
{ name: "Express", what: "checks by hand", score: 46.1, values: { hello: 44, params: 46, body: 71, answer: 72, router: 23, notFound: 22, invalid: 50 } },
{ name: "inkan", what: "on node:http", ours: true, score: 89.8, values: { hello: 90, params: 93, body: 87, answer: 88, router: 91, notFound: 95, invalid: 81 } },
{ name: "inkan, sealed", what: "inkan seal, measured with 0.9.0", ours: true, score: 88.3, values: { hello: 85, params: 92, body: 98, answer: 88, router: 89, notFound: 88, invalid: 76 } },
{ name: "Fastify", what: "schemas, like inkan", score: 88.6, values: { hello: 92, params: 90, body: 93, answer: 80, router: 94, notFound: 92, invalid: 76 } },
{ name: "Hono", what: "checks by hand", score: 85.6, values: { hello: 94, params: 77, body: 93, answer: 98, router: 77, notFound: 85, invalid: 87 } },
{ name: "Express", what: "checks by hand", score: 48.9, values: { hello: 53, params: 52, body: 69, answer: 73, router: 23, notFound: 23, invalid: 53 } },
];
Loading