yuru ai engineering
product · shipping

yume

a native desktop app for agentic coding. it does not wrap an API or reimplement a harness. it spawns the official CLI and gives it a real interface: agents running in parallel on isolated branches, the model's reasoning streamed live, and a usage guard that acts before you hit a limit instead of apologizing after. built in rust and tauri, so it stays out of your way.

what it is
a native GUI for agentic coding on top of the official CLI
built with
tauri 2 and rust for the shell, a node service for CLI integration, react for the interface
platforms
macOS 11+ (apple silicon and intel), windows 10+ (x64 and arm64), linux (x64 and arm64, flatpak)
requires
the claude CLI installed. yume drives it, and never touches your OAuth tokens.
telemetry
none. no tracking, no analytics, and no markup on your tokens.
price
free demo · $4/mo or $49 once for pro

the three things that are hard to copy

split panes and themes are table stakes. the field ships them too. these are the parts that took real work.

01
a usage guard that acts, not just warns

most tools show you a meter. yume acts on it: it declines to start a turn at 99% of your 5-hour window, and reroutes an expensive turn to a cheaper model at 98% of the weekly bucket. it reads your usage through the CLI itself, never your credentials, and fails open if it cannot.

02
it owns the permission surface

rather than inheriting a blanket bypass, yume replaces it with per-session tool control across eight categories, MCP isolation that ignores repo-local server configs, and bash deny rules evaluated before execution. the one thing it cannot do, stop a specific tool call once a turn is already running, is stated plainly in the docs rather than buried.

03
maintained against the CLI's release cadence

the underlying CLI ships constantly. yume keeps a per-version review log reaching back to CLI 2.1.113, which is the unglamorous work that decides whether the integration still functions next week.

what you get

parallel agents
four built-in roles (architect, explorer, implementer, guardian) running concurrently, each on its own git branch or worktree, following an understand → decompose → act → verify loop.
live thinking
watch reasoning stream as it happens, with an interleaved thinking toggle and five effort levels, instead of staring at a spinner.
context you do not lose
auto-compaction with a configurable threshold, tiered warnings before you hit it, mid-stream context injection, and message rollback when a turn goes wrong.
crash recovery
snapshots every five minutes with a 24-hour recovery window, so a crash costs you minutes rather than a session.
keyboard first
58 shortcuts, a command palette spanning 80+ commands, and a full vim mode with real keybindings.
not locked to one vendor
claude natively, plus gemini, openai, and kiro through a compatibility shim, so switching models is a setting rather than a migration.
extensible without coding
a plugin system, a skills system, MCP support, and a hook system covering 23 lifecycle events.
yours to configure
18 themes (12 OLED-black darks and 6 lights), custom colors and fonts, split panes, tabs and windows, and seven interface languages.
tauri 2 rust react zustand vite typescript MIT code license

pricing

every feature is available in the demo. the paid tiers lift the workspace limits, and that is the only difference.

demo
free
no account, no expiry
  • 2 tabs, 2 panes, 1 window
  • every feature unlocked
  • no telemetry
pro monthly
$4/mo
cancel anytime, access to period end
  • 99 tabs, 99 panes, 99 windows
  • all updates while subscribed
  • 7-day offline grace
yume never adds a markup on tokens. you bring your own CLI and your own plan, and pay us for the interface. the source is MIT licensed; the packaged app is licensed by tier, and paying lifts the workspace limits and funds the maintenance.

why we build a product at all

because shipping and supporting software is the fastest way to stay honest about what agentic systems can actually do. yume is built with the same agent loop it exposes, so every rough edge in agentic coding reaches our issue tracker before it reaches a client.