Archon organizes AI coding sessions by workspace, worktree, and provider. Launch with the CLI, monitor in the TUI, control from either.
[14:32:01] Analyzing auth middleware structure...
[14:32:03] Found JWT validation issue in middleware.ts:47
[14:32:05] Proposing fix: add token expiry check
[14:32:07] ⚡ Approval: edit src/middleware.ts (+12 -3)
[CLD] wants to edit:
src/middleware.ts +12 -3
1 run [CLD] Auth refactor
2 run [CDX] Fix tests
3 awt [CDX] API migration
4 run [OPN] Benchmarks
~/.archon/config.toml
[providers.codex]
command = "codex"
default_model = "gpt-5.1"
approval_policy = "on-request"
~/.archon/ui.toml
[sidebar]
expand_by_default = true
[chat]
timestamp_mode = "relative"
$ archon daemon --background
$ archon ui
✓ Daemon online at :7777
✓ TUI launched
How it works
archon daemon --background starts the local daemon at 127.0.0.1:7777. It manages all session lifecycles and provider IPC.
archon start --provider codex -- "fix auth". Each session gets a provider, title, and worktree binding. Or just open archon ui.
The TUI sidebar shows workspace → worktree → session. Stream output, compose messages, approve edits, interrupt or kill.
Free during beta. Local-first. Provider-agnostic.