Platform / EVΛL
EVΛL
The development environment that ships SLINK itself: AI agents working on the server, driven from a phone or a desktop.
How it works#
EVΛL runs Claude Code and OpenAI Codex on the server, inside the project, with the same memory and rules a human engineer reads (AGENTS.md, CLAUDE.md, decisions, open threads). You talk to a session; the agent edits, tests and deploys through a guarded pipeline.
| Piece | Detail |
|---|---|
| Sessions | Grouped by project, with branch and a running indicator. Resume a job after leaving the app; jobs keep running on the server. |
| Agents | Claude, GPT (Codex), or both. In both, GPT leads the conversation and writes the plan; Claude Code executes it and runs the tests. |
| Permissions | plan · auto · full, chosen per session. |
| Changes | git status, diff and log per turn; commit, discard or pull from the phone. |
| Code | Browse and edit project files on the server; the desktop app also opens local folders. |
| Environment | Skills, MCP servers, tools, models, work method, projects (clone from GitHub). |
Desktop#
EVΛL Desktop is an Electron app for Windows that wraps the same environment, with local folder access and a settings dialog for providers (any OpenAI-compatible endpoint), agents and projects. It updates itself.
Guarded deploys#
ops/stage.sh # unit tests → staging restart → live checks → browser tests
ops/promote.sh # merge → restart live → health check → auto-rollback on failure