Skip to content

README.md

Latest commit

 

History

History
198 lines (142 loc) · 7.79 KB

README.md

File metadata and controls

198 lines (142 loc) · 7.79 KB

🪡 RouteCode

An AI coding assistant that lives in your terminal — powered by any LLM.

CI Release License: GPL v3 Discord

Install

⚡ One-liner

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/anasx07/routecode/main/install.sh | sh

Windows (PowerShell):

irm https://raw.githubusercontent.com/anasx07/routecode/main/install.ps1 | iex

Downloads a self-contained binary and puts routecode on your PATH automatically.

🦀 Via Cargo

cargo install --path apps/cli

📦 Manual binary download

Grab the right binary from the latest release:

Platform Binary
Windows x86_64 routecode-windows-x86_64.exe
macOS Apple Silicon routecode-macos-arm64
macOS Intel routecode-macos-x86_64
Linux x86_64 routecode-linux-x86_64

Place the binary anywhere on your PATH and rename it to routecode.

Quick start

routecode                            # start interactive session
routecode --model gpt-4o             # specific model
routecode --provider openrouter      # specific provider
routecode --resume last_session      # resume last session

On first run, RouteCode will ask for your API key and save it to ~/.routecode/config.json.

Supported providers

Provider Env var Base URL (if custom)
OpenRouter OPENROUTER_API_KEY -
NVIDIA NVIDIA_API_KEY https://integrate.api.nvidia.com/v1
OpenCode Zen OPENCODE_ZEN_API_KEY https://api.opencode.ai/zen/v1
OpenCode Go OPENCODE_GO_API_KEY https://api.opencode.ai/go/v1
OpenAI OPENAI_API_KEY -
Anthropic ANTHROPIC_API_KEY -
Google GEMINI_API_KEY -
DeepSeek DEEPSEEK_API_KEY -
Cloudflare Workers AI CLOUDFLARE_API_KEY (or account_id:token) -
Cloudflare AI Gateway CLOUDFLARE_API_KEY (or account_id:gateway_id:token) -

What RouteCode can do

  • Read and write files with AI-driven tools
  • Run shell commands with captured output
  • Session management — save and resume conversations
  • Token usage tracking — monitor costs in real-time

Slash commands

Command Description
/model Switch model mid-session
/resume Resume a saved session
/sessions List all saved sessions
/clear Clear the current conversation
/help Show all commands

Configuration

All persistent settings live in ~/.routecode/config.json (created on first run). You can edit the file by hand — unknown fields are ignored, and missing fields fall back to their defaults. The most useful fields:

Field Type Default Description
model string gpt-4o Default model used at startup.
provider string openai Default provider (see Supported providers).
api_keys object {} Map of provider id → API key.
thinking_level string default One of default / low / medium / high (provider-dependent).
sub_agents_enabled bool true Allow the orchestrator to spawn sub-agents for subtasks.
retry_policy object {"strategy": "disabled"} Retry strategy for failed provider requests. Tagged shape: {"strategy": "disabled"}, {"strategy": "qir"}, or {"strategy": "exponential_backoff", "max_attempts": 5, "base_secs": 1.0, "jitter": true}. Currently only qir and disabled are honored by the orchestrator; exponential_backoff is reserved. See below.
vertex_project string "" GCP project id, required when provider = "vertex".
vertex_location string us-central1 GCP region for Vertex.
allowlist string[] [] Extra filesystem paths the bash / file tools are allowed to touch.

retry_policy strategies

strategy Behavior
disabled No retry on failure. Default.
qir Experimental. Quick Infinite Retry — re-send failed requests immediately with no delay and no attempt limit. Use at your own risk — aggressive retrying can rate-limit, suspend, or permanently ban your account with Anthropic, OpenAI, Google Gemini, OpenRouter, Cloudflare, Vertex, OpenCode Zen, etc. The RouteCode project and its authors accept no responsibility for bans, lost credits, or other consequences. Toggle this in Settings → Engine & Models → Quick Infinite Retry.
exponential_backoff Reserved for future use. The fields max_attempts, base_secs, and jitter are accepted but not yet honored at the call site.

Note: retry_policy is snapshotted at the start of each request and is NOT re-read mid-run. If you want to stop a runaway retry loop, use the Stop button (sends a cancel signal) — don't just toggle QIR off.

Migration: Old configs with quick_infinite_retry: true (or retry_policy: true) are auto-migrated to retry_policy: {"strategy": "qir"} on load. A deprecation warning is logged. The legacy field is never written back out — save once to clean up.

Building from source

git clone https://github.com/anasx07/routecode
cd routecode
cargo build --release -p routecode-cli
# Binary is located at target/release/routecode-cli (or .exe on Windows)
./target/release/routecode-cli

Architecture & Project Structure

RouteCode is a Rust workspace consisting of a CLI application and a shared SDK.

apps/cli/               # TUI application
└── src/
    └── ui/             # Ratatui-based interface
libs/sdk/               # Core logic and AI orchestrator
└── src/
    ├── agents/         # AI Provider implementations
    ├── core/           # Orchestrator and message types
    ├── tools/          # AI tools (file_ops, bash, etc.)
    └── utils/          # Storage, costs, and shared utilities

Community

Join the RouteCode Discord — where bugs get caught, features get shaped, and contributors find each other.

Discord

  • 💬 General chat and questions
  • 🐛 Bug reports and help
  • 💡 Feature requests and roadmap discussion
  • 🎉 Show and tell — share what you built

Contributing

PRs are welcome. Please open an issue first for significant changes.

cargo test            # run tests
cargo fmt             # format code
cargo clippy          # lint

License

GPL v3.0