proof · public receipts · v0.4.27

Proof.Not just promises.

Every claim on this site traces to a verifiable artifact — open source on GitHub, live in the registry, running on the public docs.

No marketing screenshots. No private demos. The compiler ships on npm and PyPI under Apache-2.0. The registry is live with public packages. The MCP tools install in one command across eleven IDEs and agent runtimes. Read the numbers, then read the source.

shipping today

The numbersbehind the loop.

Every metric on this page is generated from axint/metrics.json in the open-source compiler repo. No marketing-massaged numbers — these are the same values the test runner emits, refreshed on every release.

v0.4.27
current release

live on npm + PyPI

1,308
tests passing

1194 TypeScript · 114 Python

204
Apple-specific diagnostics

running on every check

35
MCP tools

5 built-in prompts

58
live registry packages

with source + install commands

11
IDE + agent integrations

Claude Code, Cursor, Codex, Xcode, JetBrains, +6 more

33
Xcode auto-fix rules

applied without agent intervention

26
bundled templates

intent · view · widget · app

Apache-2.0
license

open source · forever

4
Apple surfaces

App Intent · SwiftUI · Widget · App

Numbers are generated from source, packages, docs, and public release metadata. Clone github.com/agenticempire/axint, run npm test, and the test count on this page should match the output. The numbers travel with the source.

view public truth json

last generated · May 6, 2026

what's shipping

Receipts youcan verify.

Every layer of axint is publicly verifiable — compiler source on GitHub, packages on the registry, install commands in the docs. Click through any of these and inspect the artifact directly.

01 · compiler

GitHub repo is public.

v0.4.27 of the compiler is live on npm and PyPI under Apache-2.0. The repo includes 1,308 tests, 204 diagnostics, and 26 reference templates. Clone it, run the tests, read the source — every claim on this site traces back to a file in there.

02 · registry

Live packages resolve to source.

58 packages live at registry.axint.ai today. Each one ships its source, generated Swift, install command, compiler metadata, and validation history. GitHub-backed creator identity. Bundle hashes verified on every install.

03 · mcp tools

One command installs across hosts.

35 MCP tools and 5 prompts ship with the compiler. Claude Code, Cursor, Codex, Claude Desktop, Windsurf, VS Code, Xcode, JetBrains, Zed, Neovim, plus generic MCP-compliant hosts. Drop in the install command, your agent is calling axint immediately.

04 · docs

Docs are public and crawlable.

Quickstart, fix packets, cloud check loop, MCP host setup for every supported runtime, full CLI reference, full diagnostics catalog. The docs are how the agent installs and how a human verifies that the install worked.

production usage

The compilergets better every release.

We dogfood axint in production. That's how the loop tightens. Real Apple-native features generated by real coding agents, hitting real Apple contract surfaces — and every gap that surfaces becomes a diagnostic, an auto-fix, or a tighter repair prompt in the next release.

01 · feedback signals

Real Apple-native software, end-to-end on axint.

We dogfood the compiler by building production Apple software on it. Every compiler gap, validator miss, or repair-prompt failure caught while shipping real features becomes a feedback signal that ships in the next axint release. The compiler gets better every release because real usage stress-tests it under contract-shape pressure.

02 · release cadence

Public changelog, traceable improvements.

Every release ships with a public CHANGELOG entry naming what improved and what failure mode it fixes. Read the diff between two versions and you can see the exact diagnostic that landed, the exact template that got refined, the exact MCP tool that got hardened.

03 · the loop in motion

Generate, fail, repair, ship.

The Fix Packet contract is exercised on every real run. When the agent ships clean, the loop closed; when it didn't, the failure surfaces in our internal log and the compiler gets a new diagnostic, a new auto-fix rule, or a tighter repair prompt. The loop is the moat — and the moat compounds with every shipped fix.

The flywheel: production usage produces feedback signals → feedback signals improve the compiler → better compiler attracts more usage → more usage produces more signals. Open source accelerates this because the source moves faster than anything closed.

verify it

Read the source.Run the loop.

Don't take the numbers on faith. Clone the repo, run the tests, install the MCP server, watch the compiler generate Swift, watch Cloud write the Fix Packet.

Every claim on this page is testable in under five minutes with a clean shell and node installed.