# Wilco

> A multi-model terminal coding agent that lets you switch between Claude, GPT and Grok models on one shared subscription allowance.

Wilco is a terminal coding agent that reads project files, edits code and runs commands and tests, with the model selectable mid-session across Claude, GPT and Grok families. It is sold as a seat with one shared weekly allowance rather than per-token billing. It runs on macOS, Linux and Windows 10/11 and installs with a single command.

## What It Is

The `wilco` agent is built from the open-source pi agent (MIT) with Wilco's provider, theme and defaults compiled in. Inside a project it reads and edits files and runs shell commands, and `/model` switches between supported Claude, GPT and Grok models. `wilco -p "..."` runs a single prompt and exits.

The same seat also works with the official Claude Code, Codex and Grok Build CLIs through launchers such as `wilco-claude` and `wilco-codex`. Those launchers use separate configuration, so existing logins stay untouched. Wilco states it is an independent service, not affiliated with Anthropic, OpenAI or xAI.

## How Requests Flow

Clients talk to a local daemon, `wilcod`, which forwards requests to Wilco's gateway and then to Anthropic, OpenAI or xAI. Authorization happens once in the browser; local config files hold only a loopback token. The seat credential is stored in the macOS Keychain, protected with DPAPI on Windows, or kept in a user-readable file under `~/.wilco` on Linux. Any OpenAI-compatible tool can also point at the local daemon. Wilco says it never uses code or conversations for model training, and the model providers process requests under their own data policies.

## Allowance and Maintenance

All models and entry points draw from one weekly pool that refills seven days after the window opened, with no rollover. Plans last 30 days without automatic renewal. `wilco --check` diagnoses the local proxy, plan and commands, and the install script offers `--purge` and `--uninstall` switches.

## Planned Work

Task-aware routing, independent multi-model review and expert handoff are described as planned and not available in the current release. Further providers (GLM, Kimi, DeepSeek, Qwen) and VS Code and Claude desktop app support are also listed as planned.

## Features
- Terminal coding agent that reads, edits files and runs commands and tests
- Switch between Claude, GPT and Grok models with /model
- One shared weekly allowance across all models and CLIs
- Launchers for Claude Code, Codex and Grok Build
- OpenAI-compatible local endpoint for other tools
- One-line install on macOS, Linux and Windows
- Diagnostic check, purge and uninstall switches
- Dashboard showing usage and weekly reset time

## Integrations
Claude Code, Codex, Grok Build, Anthropic, OpenAI, xAI, OpenRouter, Cline, Roo

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, VSC_EXTENSION, CLI

## Pricing
Paid

## Links
- Website: https://wilco.run
- Documentation: https://wilco.run/docs
- EveryDev.ai: https://www.everydev.ai/tools/wilco