RAYU v2.0.1

The Autonomous AI Coding Agent Built for Software Engineers.

Connect any LLM, deploy self-healing multi-agent swarms, orchestrate 600+ MCP tools, and inspect deterministic diffs — right inside your terminal. Zero vendor lock-in. Zero code telemetry.

$curl -fsSL https://rayucode.com/install | bash
Standalone binary · Self-updating · ~3s installinspect script →
AK
SR
MJ
TL
Trusted by 2,400+ engineers shipping agents daily
// Rayu Agentic Runtime ready · Type commands below or let demo run
>rayu "fix the auth bug"
●Read(src/auth.ts)
Reading src/auth.ts…
142 lines read
●Edit(src/auth.ts)
- if (now > expires) {
+ if (now < expires) {
●Bash(bun test)
$ bun test
✓ 12 tests passed in 0.41s
✦Done. Fixed token expiry check on line 47.
>▒
// Orchestrate Across 20+ Frontier & Local Runtimes
◈Claude 3.7 Sonnet
◈DeepSeek R1 Reasoning
◈GPT-4o
◈Gemini 2.0 Flash
◈Llama 3.3 70B
◈Qwen 2.5 Coder
◈Mistral Large 2
◈Ollama Local
◈Groq LPU
◈Model Context Protocol (MCP)
◈AWS Bedrock
◈Google Vertex AI
◈Together AI
◈OpenRouter
◈Cerebras Fast Inference
◈NVIDIA NIM
◈vLLM Self-Hosted
◈Claude 3.7 Sonnet
◈DeepSeek R1 Reasoning
◈GPT-4o
◈Gemini 2.0 Flash
◈Llama 3.3 70B
◈Qwen 2.5 Coder
◈Mistral Large 2
◈Ollama Local
◈Groq LPU
◈Model Context Protocol (MCP)
◈AWS Bedrock
◈Google Vertex AI
◈Together AI
◈OpenRouter
◈Cerebras Fast Inference
◈NVIDIA NIM
◈vLLM Self-Hosted
◈Claude 3.7 Sonnet
◈DeepSeek R1 Reasoning
◈GPT-4o
◈Gemini 2.0 Flash
◈Llama 3.3 70B
◈Qwen 2.5 Coder
◈Mistral Large 2
◈Ollama Local
◈Groq LPU
◈Model Context Protocol (MCP)
◈AWS Bedrock
◈Google Vertex AI
◈Together AI
◈OpenRouter
◈Cerebras Fast Inference
◈NVIDIA NIM
◈vLLM Self-Hosted
20+
Providers Supported
Claude 3.7, DeepSeek, GPT-4o, Ollama
600+
MCP Tool Protocols
Postgres, GitHub, Slack, Docker, AWS
<450ms
First-Token Latency
Ultra-low latency edge orchestration
0%
Code Retention
Zero training telemetry · 100% BYOK
// ARCHITECTURAL SPECIFICATION

Engineered for Deterministic Execution

How Rayu parses codebases, orchestrates specialized models, and applies verified diffs without hallucinating breaking regressions.

pipeline/01_execution.log
STAGE VERIFIED
// Stage 01: Context Engine & AST Indexing
$ rayu index --depth=full
✓ 1,420 symbols indexed
✓ Dependency graph cached
✓ Git status: 3 staged, 1 modified
State: Deterministic · Zero Hallucination GatePass 100%
// CAPABILITIES

Built for Terminal Warriors & Senior Engineers

Every feature designed to eliminate context-switching and accelerate your delivery cycle.

01 / MULTI-PROVIDER/model

Multi-Model Hot Swapping

Switch between Claude 3.7, DeepSeek R1, GPT-4o, and Gemini mid-task without resetting context. Bring your own keys or use Rayu-hosted credits.

$ rayu /model deepseek-r1
02 / PROTOCOL600+ Tools

Model Context Protocol (MCP)

Native JSON-RPC client for Model Context Protocol servers. Connect databases (Postgres, Redis), GitHub PRs, cloud providers, and browser automation.

mcp.call("postgres/query")
03 / SWARMSParallel Agents

Parallel Agent Swarms

Spawn subagents that work concurrently on frontend UI, backend endpoints, and test suites with shared memory and conflict resolution.

$ rayu spawn --agents=3
04 / TEST LOOPSAuto-Healing

Self-Healing Test Cycles

Runs pytest, bun test, or cargo test automatically. Captures compiler errors, diagnoses failures, and patches code until all tests pass 100%.

✓ 47 tests passed in 0.8s
05 / PRIVACYAir-Gapped

Local & Air-Gapped Modes

Run 100% offline with local Ollama, LM Studio, or vLLM instances. No proprietary code ever touches a public cloud server.

$ rayu /connect ollama
06 / MEMORYStateful

Persistent Workspace Memory

Rayu records project architecture rules, custom linters, and past bug resolutions in local memory files, retaining context across team sessions.

.rayu/memory.sqlite cached
// BENCHMARK COMPARISON

Why Rayu Outperforms Traditional CLI Agents

Engineered from bare metal for zero lock-in, extreme speed, and full Model Context Protocol standard compliance.

Capability / Metric◈ RAYU CLIClaude CodeOpenCodeCursor CLIGitHub Copilot
Response Latency<450ms (Edge Route)~3–6s~4–8s~2–4s~1–3s
Multi-Provider BYOK20+ Providers & OSSAnthropic OnlyLimited (3-4)Proprietary ProxyOpenAI / Claude Only
Local / Air-Gapped ModelsNative (Ollama / vLLM)No (Cloud only)ExperimentalNo (Cloud only)No (Cloud only)
600+ MCP ToolsNative JSON-RPC 2.0PartialNonePartialNone
Parallel Agent SwarmsNative OrchestrationSingle ThreadSingle ThreadSingle ThreadSingle Thread
Telemetry & PrivacyZero Code RetentionCloud DependentPrivateCloud TelemetryCode Telemetry
Core LicensingMIT Open SourceProprietary ClosedOpen SourceProprietary ClosedProprietary Closed
// AVAILABLE THROUGH THE RAYU API

Live Hosted Model Catalog

Copy a Rayu model ID directly into your API request and compare its current credit charges and capabilities.

This lists every globally enabled model. Your plan and API-key scope determine which models your key can invoke.

// TESTIMONIALS

Built & Praised by Real Software Engineers

Real feedback from engineering leaders, founders, and students who ship code with Rayu CLI every day.

★★★★★
◈ Multi-Agent Routing

“What sets Rayu apart from OpenCode and the rest of the open-source field is per-agent model assignment — Gemini on frontend, Claude Sonnet on backend, DeepSeek R1 running the orchestrator.”

LK
Ly Yeak Khai
CEO & Founder · Senior Software Engineer · Dockified
★★★★★
◈ Engineering Pride

“It's cool that Cambodia gets to own a tool like this, instead of just reaching for Claude Code every time. The local CLI performance is unbelievable.”

MM
Mengsry Mey
Senior Software Engineer · MRTJC
★★★★★
◈ Zero Regressions

“Genuinely fun to use, and faster than Claude Code or OpenCode at getting to a good result. The deterministic AST diffs save us hours every sprint.”

KP
Khenchandara Pisey
Senior Web Developer
★★★★★
◈ MCP Tooling

“It ships image and video generation built in, plus 600+ MCP tools, so our frontend engineers can stay inside one terminal for the whole development cycle.”

SE
SeavFou Eang
Software Engineering Student · CADT
// FAQ & SPECS

Frequently Asked Questions

Technical specifics for developers integrating Rayu into their workflows.

READY TO REVOLUTIONIZE YOUR WORKFLOW

Your Next Pull Request is One Command Away.

Install the pre-bundled standalone binary in seconds. Bring your own model keys or explore Rayu hosted credits with zero commitment.

Open Source MIT◈macOS, Linux & Windows◈Node Bundled