Rebon Core v0.9.2 (rustc 1.79.0, x86_64) — tty1 login: rebon@local
██████╗ ███████╗██████╗  ██████╗ ███╗   ██╗
██╔══██╗██╔════╝██╔══██╗██╔═══██╗████╗  ██║
██████╔╝█████╗  ██████╔╝██║   ██║██╔██╗ ██║
██╔══██╗██╔══╝  ██╔══██╗██║   ██║██║╚██╗██║
██║  ██║███████╗██████╔╝╚██████╔╝██║ ╚████║
╚═╝  ╚═╝╚══════╝╚═════╝  ╚═════╝ ╚═╝  ╚═══╝
multi-agent coding engine · rust-native · local-first
last login: Fri Jul 3 09:12:04 2026 on tty1
rebon@local~/workspace
[]runtimestarting…
[]sandboxstarting…
[]plannerthinking…
[]coderthinking…
[]reviewerthinking…
[]runnerthinking…
LAUNCHING INTERFACE0%
● BOOT · rebon@local · tty1 · 6 tasks 0%
Rebon CODE AI
Docs Get started
RUST-NATIVE · MULTI-AGENT · LOCAL-FIRST

Multi-agent coding, engineered in Rust.

Rebon Code AI orchestrates specialized agents inside a high-performance Rust runtime — with a fully local Agent Studio and sandboxed execution. No cloud round-trips. No black boxes.

Start building → See it work
12×
faster tool calls
0
cloud round-trips
100%
local sandbox
4+
parallel agents
SCROLL
01 — CAPABILITIES

Built for engineers who ship, not for demos.

Everything runs on your hardware, at native speed, with the determinism a real codebase demands.

Multi-Agent Orchestration

Specialized agents — planner, coder, reviewer, runner — coordinate through a shared task graph. Parallel by default, deterministic by design.

Rust-Native Performance

A zero-GC runtime with fine-grained scheduling. Context assembly, tool calls and file I/O measured in microseconds — not seconds.

Local Agent Studio + Sandbox

Author, run and inspect agents entirely on your machine. Every command executes in an isolated sandbox with policy-gated shell and network.

Streaming diffs Deterministic replay MCP tools Git-aware context Context caching Policy-gated shell
02 — HOW IT WORKS

A local pipeline, from prompt to passing tests.

The Rust orchestrator drives every stage. Nothing leaves your machine unless you say so.

STUDIO

Agent Studio

Author, run & inspect agents locally.

RUST CORE ⚡

Orchestrator

Task graph · scheduler · context engine.

AGENTS

Agent Pool

Planner · Coder · Reviewer · Runner.

SANDBOX

Isolated Sandbox

Scoped FS · shell · gated network.

OUTPUT

Your Codebase

Diffs applied, tests run, nothing leaks.

03 — PLAYGROUND

One prompt. Four agents. Zero context switches.

rebon — agent studio
rebon ❯ add token-bucket rate limiting to the REST API
▸ planner  decomposed task → 4 subtasks
▸ coder     editing src/middleware/rate_limit.rs
▸ coder     editing src/router.rs
▸ sandbox  cargo test — running…
✓ sandbox  42 passed · 0 failed  (1.8s)
✓ reviewer approved · diff ready
+ pub struct TokenBucket { capacity: u32, refill: Duration }
+ impl Layer for RateLimit { /* … */ }
- // TODO: rate limiting
ACTIVE AGENTS
Planner
decompose & sequence
idle
Coder
write & edit files
idle
Runner
exec in sandbox
idle
Reviewer
verify & approve
idle