← Back to Discover

Sillage

Self-hosted web app for running Claude Code, Codex, OpenCode and Grok Build from a browser or phone, with saved conversations, a repository panel and scheduled tasks.

Coding agentsWorkflow automation

What it does

Sillage is a self-hosted, mobile-first web interface that runs the vendors' own agent command-line tools (Claude Code, Codex, OpenCode and Grok Build) on a machine you control, replacing the terminal with a browser UI. The README explains the design choice: agents work best inside the harness their vendor ships, so Sillage drives those tools instead of rewriting them. The server keeps the event journal, so a conversation continues after the browser closes and can be reopened from a phone. It adds a message queue, full-text search across conversations, a file explorer with editor, diffs, branch graph and terminals, per-turn checkpoints that can be reverted, worktree handling, and a board connected to GitHub issues. Sessions can message one another, and Sillage supports scheduled tasks, webhook-triggered sessions, a Decisions inbox for autonomous work, shared instructions and memory across the four tools, a skill library and an MCP server. It installs as a PWA with notifications and offers a task API with scoped bearer tokens. Install options are Docker, Helm on Kubernetes, a Linux or macOS install script, or Windows through WSL2. The README says it is built for a trusted circle and should never be exposed directly to the internet.

Capabilities

  • One interface for Claude Code, Codex, OpenCode and Grok Build
  • Server-side sessions that keep running after the browser closes
  • File explorer, editor, diffs, branch graph and terminals beside the chat
  • Per-turn checkpoints that can be reverted, plus worktree management
  • Scheduled tasks and webhook automations
  • Installable PWA with push notifications

When to use it

  • Running coding agents from a phone or another device while the work happens on your own server
  • Using several coding-agent command-line tools with shared history, search, instructions and memory
  • Starting agent runs on a schedule or in response to issues, pull requests or CI failures