IT insights

PhotoCraft: the Rust Photoshop alternative built for local, scriptable editing

How PhotoCraft rebuilds Photoshop-style PSD editing in Rust, runs locally, exposes CLI and MCP automation, and where its early-alpha limits still matter.

PhotoCraft: the Rust Photoshop alternative built for local, scriptable editing article cover

PhotoCraft is a clean-room, open-source reimplementation of Photoshop-style image editing written in Rust. The project is interesting for more than its interface: it combines layered PSD editing, a GPU compositor, command-driven automation, a CLI and an MCP server in one local engine. That makes it relevant to designers, but also to developers building repeatable image pipelines or agent-controlled creative workflows.

The weekly project snapshot supplied to ITHub records a gain of approximately 17,000 GitHub stars. At the October 9, 2026 source check, the official PhotoCraft repository showed 26,943 stars. The weekly change is the supplied ranking figure; the total was verified directly through GitHub.

What is PhotoCraft?

PhotoCraft is a native image editor from the ArtCraft community. Its goal is familiar Photoshop behavior without proprietary Adobe code, assets or shaders. The maintainers describe the implementation as clean-room work based on public specifications and observed behavior. The code is available under Apache 2.0 and MIT licenses.

The application covers the workflow developers and technical designers expect from a serious raster editor: layers, groups, masks, adjustment layers, smart objects, filters, editable type, vector shapes, selections, paths, color management and multiple blend modes. It reads and writes PSD and PSB files alongside formats such as PNG, JPEG, TIFF, WebP, SVG, OpenEXR and AVIF.

That breadth explains the attention, but the architecture is the more durable reason to watch the project.

A Rust engine instead of a web wrapper

PhotoCraft is organized around a data-oriented document model and a command engine, with the user interface kept as a thinner layer. Its repository describes 24 Rust crates and two compositors: a CPU compositor used as a reference and a GPU path based on wgpu. The GPU layer can target Metal, Vulkan, DirectX 12 and WebGPU.

Images are stored in sparse 256 by 256 pixel tiles with copy-on-write behavior. That design reduces the cost of undo history and avoids copying an entire large canvas when a small region changes. Effect results can be cached against layer state, which matters when filters and adjustments remain editable.

For engineering teams, this is a better foundation than an interface that merely imitates Photoshop. The document model, renderer and commands can be tested without clicking through the desktop UI. PhotoCraft also compiles its engine to WebAssembly, although the native application remains the primary experience.

PSD compatibility is measurable, not a slogan

The project publishes concrete compatibility claims. Its README reports that reopening and saving preserves the rendered result for 307 of 309 files from the psd-tools test set and 169 of 170 files in a mixed ag-psd/psd-tools corpus. Unknown raw blocks and descriptors are carried through when the higher-level model does not understand them.

Those numbers come from the project’s own corpus and test harness, so they are evidence to examine rather than a guarantee for every production file. A Photoshop workflow can depend on plug-ins, uncommon typography, linked assets, color profiles or undocumented PSD behavior. Test representative documents, compare the composite output and retain the originals before adopting any alternative editor.

PhotoCraft also tests CPU and GPU rendering against each other and reports more than 1,700 tests across format round trips, compositing and multi-depth behavior. The source tree exposes the corpus tooling, which makes failures easier to reproduce than a closed compatibility claim.

Why the command system matters

Every action is represented as a command. The UI, CLI, JSON control channel and MCP server call the same registry, which the project says contains more than 500 commands. A developer can open a PSD, apply several operations and save the result without automating mouse clicks:

photocraft-cli run input.psd --cmd filter.sharpen.smartSharpen --params '{"amount":80}' --out output.png

The same model supports batch execution over a folder. For an AI agent, the MCP endpoint exposes the editing engine as explicit operations instead of asking a vision model to guess where a button is. PhotoCraft also offers an authenticated loopback control channel for inspecting and driving a running desktop session.

This is the point of contact with the wider developer tools directory: PhotoCraft can act as a creative application and as programmable infrastructure. Automated product screenshots, asset conversion, PSD validation and repeatable export recipes are plausible uses, provided the team validates output quality.

Features developers should evaluate first

PSD round trips. Build a sample set containing the exact layer effects, masks, color modes and fonts used by your team. Save each file and compare both the composite image and editable structure.

Color management. PhotoCraft supports RGB, Grayscale, CMYK and Lab documents at multiple bit depths, with ICC profile assignment, conversion, soft proofing and gamut warning. Print and photography teams should still run device-specific tests.

Automation stability. Treat commands as an API. Pin the PhotoCraft version, store the command list with your pipeline and compare generated output in CI before updating.

Resource behavior. Large documents, smart objects and filter stacks can stress memory and GPU drivers. Measure the documents that matter on the hardware that will run them.

Installation and supported systems

The project publishes signed or packaged builds for Windows, macOS, Linux and FreeBSD, plus a WebAssembly build. Windows releases include MSI and portable archives; macOS uses a signed and notarized universal build; Linux options include AppImage, Flatpak, Debian, RPM and tar archives.

Developers can build from source with Cargo:

git clone https://github.com/storytold/photocraft
cd photocraft
cargo run --release -p photocraft -- image.psd
cargo test --workspace

Use the release checksums and the project’s security documentation when distributing builds inside an organization.

The limits are significant

PhotoCraft is early alpha. Its maintainers explicitly say it is not yet a daily professional Photoshop replacement. Major gaps include generative AI features, approximately twenty missing tools, typography depth, advanced professional workflows and plug-in compatibility. A wired menu command also does not prove that every edge case behaves like Photoshop.

That honesty is useful. PhotoCraft already has enough surface area for testing, automation experiments and some real editing work, but migration should be gradual. Keep Photoshop or another established editor available for documents that fail your acceptance suite.

Who should try PhotoCraft?

PhotoCraft is most compelling for Rust developers, open-source contributors, privacy-sensitive creative teams and engineers who need a scriptable PSD engine. It is also worth evaluating if local processing and inspectable file handling matter more than Adobe ecosystem integration.

Studios dependent on proprietary Photoshop plug-ins, exact typography or mature generative features should treat it as an experiment. The right proof of concept is a folder of difficult production documents, not a blank canvas demo.

Verdict

PhotoCraft is one of the more technically ambitious open-source creative applications to appear in 2026. Rust and GPU acceleration attract attention, but the command architecture and published PSD corpus results make the project substantively different. It is not finished, and its README says so. For developers, that makes the opportunity clear: evaluate the engine now, contribute reproducible failures and adopt only the workflows that pass your own tests.

Official sources and review

Reviewed by Emanuel DE ALMEIDA on October 9, 2026. The main references were the official repository and README, its architecture documentation, security policy and parity assessment. Project totals and capabilities can change after publication.

About the author

Emanuel DE ALMEIDA: Emanuel DE ALMEIDA is an IT journalist and editor at ITHub Directory, covering cybersecurity, cloud infrastructure, DevOps, and enterprise technology.