SpliceCraft is a plasmid workbench that runs where you already work. Open a map, edit the sequence, design primers, plan a Golden Braid or MoClo assembly, BLAST a hit, check your Sanger reads, and keep a lab notebook — all from the keyboard, all in one place, no browser tab and no cloud account. Circular and linear maps render as crisp Unicode braille graphics in any modern terminal, and nothing leaves your machine unless you ask it to.

It's built by a practicing bioengineer for daily bench work: the bug reports come from real cloning, and so do the fixes. If you live in a terminal and clone for a living, it's meant to feel like home.

Why give it a try:

• Fast and local. No Electron, no web app, no login. pipx install splicecraft and you're designing in seconds.
• It does the whole job. View → edit → design → clone → simulate → verify → document — one tool that actually understands how those steps connect.
• It guards your data like it's irreplaceable (because it is — see below).
• It's scriptable. A 100+ endpoint local API and a stdlib CLI let an agent or a shell script drive every workflow.

pipx install splicecraft
splicecraft                      # empty canvas
splicecraft L09137               # fetch pUC19 from NCBI on launch
splicecraft myplasmid.gb         # local GenBank or .dna

x86-64 Linux, Intel macOS, and Windows install entirely from prebuilt wheels — nothing to compile. On ARM64 Linux (Raspberry Pi / ARM cloud) and Apple Silicon with Python ≥3.10, one dependency (primer3-py, the primer-design engine) has no prebuilt ARM wheel and compiles at install, so install a C toolchain first: sudo apt install build-essential python3-dev (Linux) or xcode-select --install (macOS). One-time, then pipx install splicecraft.

Press ? once running for the full keyboard-shortcut reference. See docs/install.md for pip / uv / conda / source installs.

Your plasmid library is the product of months — sometimes years — of work. SpliceCraft is engineered so it can be your daily driver without ever making you nervous about it. Three promises, and the receipts behind them:

Your data is sacred. Every save runs through a four-layer safety net: an atomic write (a crash can never leave a half-written file), a .bak plus rotating timestamped backups, a daily snapshot, and a "suspicious shrink" guard that flat-out refuses to replace a 156 MB library with an empty file. Open a file that collides with something you already have and SpliceCraft asks — skip, keep a copy, or overwrite — it never clobbers your work behind your back. Before any self-update it snapshots everything to a sibling directory first, so even a hypothetical bug in a new version can't take your recovery copy with it.

The biology is correct, and we prove it. Restriction scanning handles palindromes, Type IIS enzymes, and origin-spanning cut sites the way a bench scientist expects; features that wrap the origin survive every edit; translation (including non-standard genetic codes via /transl_table), reverse-complement, and IUPAC matching are pinned down to the base. Selection markers are read from the actual entry vector you're using — no hardcoded antibiotics quietly mislabeling a construct. There are 4,000+ automated tests behind all of it, plus property-based fuzzing on the biology primitives, crash-injection tests on the save path, and concurrency fuzzing on the data layer. Every release ships only when the whole suite is green.

We go looking for trouble. The codebase is governed by a long list of "sacred invariants" (documented in CLAUDE.md), and each release is preceded by deep, multi-pass audits hunting for edge cases, data-loss windows, race conditions, and security gaps — with every finding verified against the real code before a line is changed. The short version: it's a workhorse, and it goes to real lengths to keep "it just works" true.

Full data-safety writeup: docs/data-safety.md · Security policy: SECURITY.md.

Everything lives behind a menu bar across the top of the screen. Here's the tour — starting with BLAST and working rightward across the bar, with the housekeeping menus (File, Settings) saved for last.

Search without leaving the app (Ctrl+B). The Local tab runs BLASTN / BLASTP / HMMscan against your own library with an in-process engine — powered by pyhmmer, so there's no external blast+ to install — and a one-click downloader pulls Pfam-A or NCBIfam (or any HMMER3 .hmm.gz URL) with automatic update-detection, no wget + hmmpress ritual. On native Windows, BLASTN/BLASTP run in-process but HMMscan needs WSL2 (pyhmmer/HMMER is POSIX-only). The Online tab sends DNA / RNA / protein — or a whole plasmid or a single feature straight from your library — to NCBI (blastn / blastp / blastx / tblastn / tblastx) or to Pfam at EMBL-EBI, and drops the hits into a scrollable table just like a web BLAST. A live spinner and poll counter prove the search is really running, and Cancel actually stops it.

Drive the restriction overlay: show all sites, unique cutters only, 6+ or 4+ bp recognition, or just the Golden Braid connectors. A multi-cutter wears a superscript cut-count on its name (EcoRI², BsaI³) on the map and sequence panel — live, so editing out a site ticks it down and you can confirm you killed a cut. Build enzyme collections — named subsets of the master catalog (200+ NEB enzymes plus your own customs) — and the active collection scopes every scan. Add a custom enzyme (name, site, cut positions, type, supplier) and it's live in every map from then on.

A workbench for your reusable annotations — promoters, RBSs, tags, CDSs. Capture a region off any plasmid into the feature library, then drop it back onto another construct to either annotate a selection or splice the sequence in. It's the same store the Synthesis editor and Domesticator pull from, so your parts stay consistent everywhere.

Marking things up by hand? Ctrl+F finds a DNA subsequence — fuzzy, so you can allow a few mismatches, and on both strands — then n / N step through every hit. Each match lands pre-selected, so Alt+Shift+F tags it as a feature on the spot: ideal for walking a plasmid and annotating repeat regions one by one. (Looking for an existing annotation instead? Ctrl+/ searches your features by name.)

A full-screen Primer3 designer covering detection, cloning, Golden Braid, and generic primers — each with a persistent Designed → Ordered → Validated lifecycle, so you can see at a glance what's been ordered and what came back working, right alongside the plasmid it belongs to.

The library beside the designer organises primers into collections (new / rename / delete, like the plasmid side, with destructive prompts defaulting to No): press Space to cycle a primer's mark — ★ select → $ cart → M move → none — then MOVE the M-marked primers to another collection, or bulk-delete / re-status the ★-marked set (the warning names exactly how many go). When it's time to order, export a collection — or your order Cart (the $-marked primers) — to a paste-ready CSV (name, sequence, length, Tm) for the synthesis company, and import a CSV back for a full round-trip. Exports refuse a malformed oligo and imports skip + report invalid rows, so a wrong sequence never reaches your order or your library.

Site-directed mutagenesis, with a hint of whimsy in the name. Point at any CDS, name the change (say, L54A), and SpliceCraft designs the SOE-PCR primers for you — with a smart fallback to a 2-primer modified-outer strategy when a near-the-end mutation can be folded into a single primer. It checks its own work, too: the shortcut is only offered when the primer genuinely carries the change, so you never amplify wild-type by accident. It also turns a pasted protein into a ready-to-order CDS — frequency-matched codon optimization against your chosen table, with a stops selector (1–3) that also honors a trailing * run when you want a double or triple stop codon, and an Avoid sites picker that scrubs your chosen restriction-enzyme cut sites out of the optimized CDS.

Its second tab, Scrub, cures a whole plasmid of restriction sites without any cloning. Pick the enzymes to remove (Golden Gate / MoClo Type IIS sites by default) and SpliceCraft finds the smallest set of point changes that destroy each site — silent ones inside any coding sequence, checked against every overlapping reading frame so a protein never changes (and biased toward your chosen codon table's frequent synonyms when there's a choice), and minimal swaps elsewhere — without ever spawning a new forbidden site. You get the cured circular plasmid (one click to apply, fully undoable) plus primers designed off the cured sequence so they bind exactly where they're drawn — save them to your primer library or add them straight to the map. Choose how it re-circularizes: QuikChange — an improved-QuikChange pair per locus, PCR → DpnI → transform (no ligase, no assembly); or Golden Braid — split the plasmid at each cure into BsaI-tailed fragments that a one-pot Golden Gate reaction ligates back together, the native 4 nt junction overhangs making reassembly seamless so the only net change is the cured sites. (BsaI being the assembly enzyme, every BsaI site is force-cured, and the design proves itself by a digest-and-ligate simulation before you order.) Sites it can't remove silently are reported, never forced.

A gene-synthesis composer with three tabs. The DNA tab is a horizontally-scrolling linear editor with anti-parallel strand markers, live feature stripes, restriction overlay, and AA translation, plus a feature library side-pane (Insert to splice, Annotate to overlay), and a feature-aware paste — copy a stretch of a plasmid (Ctrl+C) and paste it here, and its features ride along, annotated at the right positions. The Protein tab lets you type or paste amino acids and watch the codons appear underneath using your chosen codon table — pick a different one or hit Manage to open the tabbed codon-table manager, where each way of getting a table has its own tab: Build from genome (give an NCBI assembly accession or a taxid, and choose highly-expressed genes — ribosomal proteins, the recommended bias for strong expression — or the whole genome), Fetch (Kazusa), Import TSV, plus a Library tab to pick or delete and a Chart tab that draws any table as the classic genetic-code grid — each codon annotated with its usage within its amino-acid family (relative synonymous usage), and each family's single most-used codon highlighted in bold green (family-wide, so a codon split across cells — Leu, Ser, Arg, the stops — lights up once) for easy visual identification. The same manager opens from Settings ▸ Codon Tables and the Mutato / Constructor codon pickers. Codons are reflected live — and a built-in motif library (His6, FLAG, HA, TEV, P2A, NLS, GS linkers, and ~30 more) inserts pre-colored tags. Hit Optimize → DNA to codon-optimize the protein — the Stops selector (0–3) auto-tracks the protein's trailing * run (override it and your choice wins), alongside the same Avoid sites scrubbing as Mutato — and hand the CDS straight to the DNA tab as an editable fragment. Or hit Open to load a sequence straight from a single-entry FASTA (or other amino-acid file) — a file browser highlights the loadable formats in pink. Compose a part, hit Clone Fragment, and pick how to clone it: any modular grammar (Golden Braid L0, MoClo, your own custom grammars) hands the fragment straight to the Domesticator as an L0 building block, while Gibson and Traditional (restriction / ligation) open the Constructor on the matching tab with the fragment already pasted in. Nothing is saved and nothing on your canvas is touched at this point — you name things later, at the save step. If the modular grammar you pick has no entry vector bound yet (a fresh install, say), SpliceCraft asks you to pick an acceptor plasmid from your library and checks it really carries that grammar's dropout cassette before binding it, so the clone can't quietly degrade to a bare insert. When you save the domesticated part, one dialog names — and independently files — three things: the cloned plasmid (the part actually cloned into the entry vector — a full circular plasmid — into any collection), the linear fragment you'd send for a DNA-synthesis order (its own name, defaulting to a FRAG-… form, into any collection), and which parts bin the L0 part files into. The fragment is the primed amplicon — the insert flanked by the designed domestication primers' enzyme sites + overhangs, exactly as it would run on the bench — and both the fragment and the clone carry the domestication primers, drawn with their bound (annealing + overhang) and unbound (enzyme-tail) regions, so you can see exactly how each was built (and regenerate the amplicon later for a synthesis order). The fragment also keeps whatever features you'd annotated on it (the optimized CDS bar, say), redrawn at the right bases on the amplicon. Both the DNA and Protein tabs also carry a Clear button — a mouse twin for the keyboard reset — that empties the editor, prompting first if you have unsaved edits. The Operon Design tab turns the codon optimizer and SpliceCraft's built-in RBS engine into an expression-tuning workbench. Keep a library of protein collections on the right (add a protein by pasting a sequence, grabbing a CDS from any plasmid in your library with From feature, or Fetching one from NCBI by accession), drop proteins into the assembly lane, and give each gene a target relative RBS strength. Assemble codon-optimizes every CDS and reverse-designs every ribosome binding site in its real assembled context — so the achieved strength tracks your target, and a gene the surrounding sequence can't drive strongly is flagged rather than silently missed. The finished operon (promoter + RBS + CDS per gene + terminator) drops into the DNA tab fully annotated, ready to add cut sites or Clone Fragment. It's all pure-Python — the RNA folding, cofolding, and translation-initiation model ship inside SpliceCraft with no external dependencies. Save (and Save As, which forks a copy and only lights up once the fragment has been saved once) let you pick which collection the fragment lands in, and keep editing it there.

Your Parts Bin — the Level-0 building blocks for grammar-based assembly, organized into per-grammar bins. Multiple bins live side by side as Parts Bin collections, so a yeast toolkit and a plant toolkit never get mixed up.

Where it all comes together. A multi-tab assembly bench — Traditional cloning, Gibson, Golden Braid, MoClo, or your own custom grammar — driven by a 4-source part picker. Every assembly, at every level, lands as one complete library entry (payload + overhangs + backbone) that carries every parent feature forward, so you can trace a finished L3 construct all the way back to its L0 parts right from the Library panel.

In-silico PCR and agarose gels. Pick a template from your library, run the PCR, then save the amplicon back to the library or send it to a gel lane. Gels render at 0.5–4% with a real Helling–Goodman–Boyer mobility curve; stack several amplicons side by side, save the whole gel to reload later, or cite it as &<gel> in your lab notebook.

Verify your constructs against real reads. Drop in a Plasmidsaurus .zip and walk three numbered tabs — pick the run, pick the sample, pick the target plasmid — then Align. The read lands as a colored bar on the plasmid's linear map (blue match / red mismatch / gray gap) with its name painted right onto the bar so a multi-read pile-up stays readable. Zoomed all the way out, each cell is shaded by how much of its span actually binds — solid blue where it matches, a red shade that deepens with the mismatch density, gray for gaps — so a partially-binding read reads as a blue/red/ gray patchwork and even a single-base mismatch still shows red in its region. Click anywhere on a read's bar to jump the sequence panel to that exact spot — centered and highlighted — so you can land on a misaligned stretch (or the precise base to re-edit) without scrolling; the full per-base alignment view is still a keystroke away in the Alignment Manager. Bulk auto-align matches a whole results folder against your library in one pass — and its confirm window shows each read's real identity, mismatched-base, and gap counts (computed by actually aligning, not just the name/k-mer match score), ready when the window opens so you can see how clean every read is before you commit. The Verification Report grades every construct (✓ verified / ⚠ near-match / ~ partial / ✗ divergent) in one sortable table — click a row to jump to the first variant. The Alignment Manager lists every stored alignment with its identity, mismatched-base count, and gap count, so one glance tells you how clean each read is — and an identity that isn't a true 100% never rounds up to "100%" (a single off-by-one base reads as e.g. 99.99%, not a false perfect score). The Library panel even shows a per-plasmid Seq badge so you can see what's been verified at a glance, alongside a Kind badge (○ plasmid · / fragment · ≈ amplicon · ρ protein) telling you what each entry is.

A genuine lab notebook, in markdown. A split-pane editor, entries grouped into projects (the way plasmids group into collections), and live colored cross-references — type @plasmid, !action, or &gel and double-click (or Ctrl+G) to jump straight to the source. Attach images, and spellcheck with F7 against a dictionary you can grow.

Every plasmid you build through SpliceCraft remembers how it was made — whether you cloned it via Golden Braid, traditional digest/ligation, Gibson, or PCR, or just edited and saved it. History opens with a Protocol summary — a numbered recipe that reads left → right like the bench ("assemble pProm + pCDS_GFP + pTerm into pENTR_L1 → TU_GFP ✂ Esp3I", with a symbol legend) — above a lineage tree that opens collapsed to the finished plasmid and its direct inputs and lets you drill in as deep as you like. Selecting a step shows its detail, including the primers used for a PCR. Every step is dated — its date and time (e.g. JUN 9 2026 14:30, written slash-free so it can't be misread as MM/DD vs DD/MM) sits right beside the action. A backbone or part reused across branches is shown once and then referenced, so even a multi-part Golden Braid / MoClo build reads at a glance. The same history rides along when you import or re-export a CommercialSaaS .dna file — and that file's own creation date lands on its top step. "How did I make this again?" is always one keystroke away.

The housekeeping. File opens local files, fetches from NCBI, saves, exports (GenBank / FASTA / GFF3), bulk-imports a folder, and restores from backup. Every GenBank SpliceCraft writes (library entries, exports, autosaves) records a Created by SpliceCraft v… on … line in its COMMENT, so a file's origin is always traceable. It also runs record-level jobs — including Clone selected region (also Alt+Shift+P): highlight any stretch of DNA in the sequence panel and name the amplicon. The cut-site dialog steers you to a working enzyme pair — it marks any enzyme whose site falls inside your selection (✗) or that's Type IIS, lists the usable ones first, and pre-picks a viable directional pair; pick a destination vector there too and it flags enzymes that vector can't be opened with (⚠), refines the suggestion to a pair that cuts both, and pre-loads that vector into the Constructor as the backbone. SpliceCraft then designs the cut-site-tailed PCR primers, builds the amplicon — carrying every annotation your highlight spans, even when the highlight crosses the plasmid origin — saves the named amplicon to your library as a reusable linear PCR product (its features and both run primers travel with it, the primers drawn as primer-binding sites), and opens the Constructor's Traditional cloning tab (pre-loaded when you chose a vector — just Simulate). Simulating digests the amplicon and purifies away the off-cuts — exactly like cutting and gel-purifying at the bench — so the recognition site reforms at each junction and none of the primer pad / outside- the-cut bases carry into the construct (no stray extra bases in your clone). The cloned product keeps those insert features, shows where the cloning primers bind (the unbound 5′ enzyme-site tail drawn as a flap), and flags any vector feature your cut sites land inside as (disrupted), so dropping an insert into a lacZα MCS reads as the knock-out it is. It's also where you Migrate Data — package your entire setup (library, collections, parts, primers, features, grammars, codon tables, settings, lab notebook, and full construction history) into one portable, compressed .zip and import it into another install, so a fresh machine picks up exactly where you left off (the import snapshots your current data first and verifies every file's checksum before replacing it, and refuses anything that isn't a genuine SpliceCraft archive). And it's home to Master Delete, a triple-gated full wipe for when you genuinely want a clean slate (typed YES, a default-No confirm, and a cool-down on the button; no shortcut, no API). Settings collapses every toggle (restriction overlay, primer-binding length, and more) into one dialog, with launchers for the grammar, entry-vector, enzyme-collection, and codon-table editors.

Want to drive all of this from a script or an agent? There's a 100+ endpoint localhost JSON API (splicecraft --agent) and a stdlib-only CLI sidecar (splicecraft-cli) — see docs/agent-api.md and docs/cli.md.

Full feature reference: docs/features.md.

python3 -m pytest -n auto -q                  # full suite (~5–6 min on 8 cores)
python3 -m pytest tests/test_dna_sanity.py    # biology correctness only (< 2 s)
python3 -m pytest tests/test_perf_regression.py  # perf gates (~3 s)

All tests run offline against synthetic SeqRecords and monkeypatched data paths; the autouse _protect_user_data fixture in tests/conftest.py guarantees no test can write to real user files.

SpliceCraft is actively maintained. The maintainer is a practicing bioengineer running real cloning workflows in it daily; releases typically go out the same week a problem surfaces at the bench. Issues and PRs welcome at github.com/Binomica-Labs/SpliceCraft/issues.

See CONTRIBUTING.md before opening a non-trivial PR — it walks through the sacred invariants, the test cadence, and the security-sensitive code surfaces.

MIT