The power user's update - Loach v1.2.0 with built-in tools, slash commands, snippet variables, context management and much more!

[ztcs-software]

If you spend a lot of time in Loach, this one's for you. v1.2.0 is the biggest update yet: drive a chat with slash commands, let models reach for new built-in tools (one of them spits out PDFs), fork a conversation before it wanders off somewhere you can't get back, manage your context window, set snippet variables, and watch code render live and open it straight in VS Code - plus much more.

👉 Download Loach v1.2.0
or update directly from the app in Settings → Updates.

Built-in tools

A set of local utilities the model can call mid-answer, each with its own toggle in Settings → Tools and all off by default: calculate, datetime, count, hash, uuid, base64, json, unit_convert, diff_text, sort, and ip. They run entirely on your machine and work inside Private Chats too. When a model reaches for tools, the calls collapse into a single "Called N tools" header above the answer - expand it to see each call's name, arguments, and result.

PDF generation

Ask a model to write something up and it can hand you back an actual document. The new built-in pdf tool turns a structured spec - headings, paragraphs, lists, tables, page breaks - into a real PDF that lands in the chat as a previewable, savable attachment.

Slash commands

Type / at the start of the composer and a command palette pops up. Run chat actions without touching the mouse: /fork, /regenerate, /compact, /private, /copy, /export, /stats, /model, /persona, /snippet, /instructions, /remember, /fetch, /list, /settings and more. Destructive ones like /clear and /delete ask first, and anything that isn't a real command is just sent as a normal message so a prompt that happens to start with / is never swallowed. The full list of slash commands can be viewed by typing /help.

Custom snippet variables

Snippets can now take {{PLACEHOLDER}} variables. Define reusable static globals once in the Snippets library, or leave placeholders blank to be prompted for values when you run the snippet.

Context management

A slim bar above the composer shows how full the model's context window is, with a popover that breaks it down into system prompt, history, and attachments. When a chat gets long, the Compact context button or the /compact command summarizes the older turns into the system prompt to free up room. The original messages stay in the transcript for scrollback behind a compaction divider - they're just dropped from what the model sees next turn. Plus, context export got a Compact context toggle that writes a summarized version instead of the verbatim transcript, and imported context can now be hidden from the transcript - folded into a single collapsed card while still being sent to the model.

Fork chats

Branch any conversation into a fresh copy - from the header, from a message's "…" menu (forks up to that point), or with /fork. The fork keeps the same model and Space and wears a "Forked from …" badge that jumps you back to the original.

Code canvas upgrades

The canvas is finally resizable, updates live as a reply streams, and gains an Open in VS Code button.

Other minor changes

• Images returned by MCP tools are no longer dropped - they surface as an inline attachment on the reply.
• Importing context keeps system turns (brought in as user turns so the model still sees them) and no longer mistakes Markdown headings for role markers.
• Hardened web-fetch SSRF screening - closed IPv6 gaps in the private/loopback/link-local classifier, and redirects are re-screened on every hop.
• Polish: corrected a stale password hint, fixed the Thinking toggle showing the wrong state, refreshed out-of-date PDF-preview wording, and tidied inconsistent archive-label casing.

👉 View the full changelog in GitHub Releases