Skip to content

an open-source alternative to the dots, bots and muses of the world, run an office of claude code/codex like agents on your laptop, sandboxes or anywhere, uses your existing subscriptions

fix(cli-agents): auto restart-and-continue after first-time CLI install (ondev-a)

Chaitanya Girice1b6c1e18e2
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 9d619115..f4ad2623 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -55,6 +55,7 @@ sleep-frozen message router and Codex workers' filesystem permissions.
### Fixed
- **Codex hive workers get full filesystem + auto-approval from spawn (parity with Claude).** A Codex-engine agent in auto mode launched with `-a never -s workspace-write`, whose sandbox scopes writes to the PTY cwd (the user's project). But a hive worker must also write to its agent folder at `<harnessHome>/hive/agents/<id>/` (move `inbox/` → `.done/`, append `memory.md`, drop outbox JSON, write deliverables) — a **different path tree from cwd**, which `workspace-write` blocked. So a freshly spawned Codex worker couldn't complete HIVE PROTOCOL housekeeping and reported "it does not have permissions … grant write permission to the agent folder." Codex's auto-mode flag is now `--dangerously-bypass-approvals-and-sandbox` — the documented equivalent of Claude's `bypassPermissions` / Antigravity's `--dangerously-skip-permissions` (skip all approval prompts **and** drop the OS sandbox), so a Codex worker has the same filesystem access and auto-approval as a Claude worker from the get-go (`src/shared/agentProvider.ts`; reference/copy updated in `src/shared/codexCommands.ts`, `OnboardingWizard.tsx`, `renderer/store/config.ts`). Claude/agy/antigravity behavior is unchanged.
- **Re-arm the hive message router on wake (god→worker delivery survives sleep).** The outbox→inbox router is a `setInterval` (`hive.routeOnce` every ~1.5s) that, like the always-on beats, freezes during true macOS system sleep. `onSystemResume()` already re-armed the mission scheduler, the fleet/breaker beats, and keep-awake on `powerMonitor` `resume`/`unlock-screen` — but it never re-armed the router. So after a long sleep (e.g. laptop closed overnight) the scheduler→god path recovered while **every agent's outbox silently stopped draining**: god→worker, worker↔worker, and broadcast mail piled up undelivered, and no `message` event was logged. The resume handler now re-arms the router (clear-then-set, idempotent) **and** immediately drains the accumulated backlog instead of waiting for the first post-wake tick; the renderer's idle inbox-wake nudge then wakes each parked recipient once its mail lands (`src/main/index.ts`). Verified by `scripts/verify-keepalive-catchup.mjs` (now also reproduces the pre-fix backlog stall and proves the re-arm + flush).
+- **Auto restart-and-continue after a first-time engine-CLI install (no dead-end).** When an agent's engine binary (OpenCode/Crush/pi.dev/Codex/…) wasn't installed, the missing-CLI short-circuit ran the provider's installer in the PTY, then printed *"click restart & continue to launch the agent"* — but no such button exists for a not-yet-started agent, so the PTY just sat at `process exited (code 0)` and the agent dead-ended. Now, on a **clean install exit**, the PTY-exit handler auto restart-and-continues: it re-runs the *same* spawn into the *same* pty/window (carrying a `noAutoInstall` flag) so the freshly-installed CLI launches with no user click, and the renderer re-arms that terminal in place (clears the "process exited" line, re-enables input) via a new `pty:relaunch` signal. Provider-agnostic (every engine's installer path) and idempotent by construction — `noAutoInstall` guarantees the installer can never fire twice, and providers with no bundled installer (manual-hint-only) are never armed for relaunch. The install banner copy is now honest ("Installed — launching the agent…") (`src/main/index.ts`, `src/main/pty.ts`, `src/preload/index.ts`, `src/renderer/src/components/terminalPool.ts`).
## [0.3.0] — 2026-06-21
diff --git a/src/main/index.ts b/src/main/index.ts
index 441b97a1..7d6bfc39 100644
--- a/src/main/index.ts
+++ b/src/main/index.ts
@@ -64,6 +64,12 @@ const ptyManager = new PtyManager();
/** Live PTY id → its hive agent id, recorded at spawn. The pty:kill handler only
* gets the PTY id, so this lets a closed tab archive the right registry agent. */
const ptyToAgent = new Map<string, string>();
+/** PTY id → the spawn it should auto restart-and-continue into once a first-time
+ * CLI install finishes. The missing-CLI short-circuit runs the engine's installer
+ * in this PTY; when it exits cleanly the exit handler re-runs the SAME spawn (with
+ * install disabled) so the freshly-installed CLI launches in the SAME pty/window —
+ * no user click. Cleared the moment it's consumed, so it can never loop installs. */
+const pendingInstallRelaunch = new Map<string, { opts: AgentSpawnOptions; owner: Electron.WebContents | null; bin: string }>();
const hive = new HiveManager(
() => readConfig().harnessHome,
(channel, payload) => {
@@ -352,8 +358,29 @@ function removeWorkerScratch(workerId: string): void {
try { rmSync(dir, { recursive: true, force: true }); }
catch (e) { console.error('[worker] removeWorkerScratch failed:', e); }
}
-// A natural PTY exit must run the same teardown as an explicit kill.
-ptyManager.setExitHandler(teardownPty);
+// A natural PTY exit must run the same teardown as an explicit kill — EXCEPT when
+// the PTY was the missing-CLI installer: a clean exit there means the engine CLI was
+// just installed, so auto restart-and-continue by re-running the SAME spawn into the
+// SAME pty/window (no user click). Provider-agnostic. Idempotent by construction: the
+// relaunch carries `noAutoInstall`, so the installer can never fire (let alone loop) a
+// second time — a binary that's somehow still missing just spawns and exits normally.
+ptyManager.setExitHandler((id, exitCode) => {
+ const pending = pendingInstallRelaunch.get(id);
+ if (pending) {
+ pendingInstallRelaunch.delete(id);
+ if (exitCode === 0) {
+ // Re-arm the renderer's pooled terminal (clear the "process exited" line +
+ // re-enable input) so the freshly-spawned CLI paints onto a clean, typeable
+ // grid, then re-run the normal spawn — which now finds the installed binary.
+ const wc = (pending.owner && !pending.owner.isDestroyed()) ? pending.owner : liveWebContents();
+ try { wc?.send(`pty:relaunch:${id}`); } catch { /* window gone */ }
+ void spawnAgentCore({ ...pending.opts, noAutoInstall: true }, pending.owner);
+ return; // an install PTY has no agent/worktree to tear down
+ }
+ // Non-zero exit = install failed; leave its honest manual-fix message on screen.
+ }
+ teardownPty(id);
+});
/** Keep the system from suspending the harness while agents are running.
* Windows Modern Standby suspends desktop apps (and their child `claude`
@@ -1654,13 +1681,13 @@ function buildMissingCliScript(bin: string, provider: AgentProvider): string {
'echo.',
cmd,
'echo.',
- 'echo If it succeeded, click restart ^& continue to launch the agent.',
- 'echo If it failed, run the command above manually, then restart ^& continue.'
+ 'echo [done] If it succeeded, the agent launches automatically.',
+ 'echo If it failed, run the command above manually, then restart the agent.'
);
} else {
parts.push(
`echo No bundled installer for the ${label} provider.`,
- 'echo Install it manually, then click restart ^& continue.'
+ 'echo Install it manually, then restart the agent to launch it.'
);
if (docs) parts.push(`echo Docs: ${docs}`);
parts.push(`echo ${rule}`);
@@ -1689,18 +1716,18 @@ function buildMissingCliScript(bin: string, provider: AgentProvider): string {
`__clirc=$?`,
`echo ''`,
`if [ $__clirc -eq 0 ]; then`,
- ` echo ' [done] Installed. Click restart & continue to launch the agent.'`,
+ ` echo ' [done] Installed — launching the agent…'`,
`else`,
` echo " [x] Install exited with code $__clirc — finish it manually:"`,
` echo ' ${cmd}'`,
...(docs ? [` echo ' Docs: ${docs}'`] : []),
- ` echo ' Then click restart & continue.'`,
+ ` echo ' Then restart the agent to launch it.'`,
`fi`
);
} else {
lines.push(
`echo ' No bundled installer for the ${label} provider.'`,
- `echo ' Install it manually, then click restart & continue.'`,
+ `echo ' Install it manually, then restart the agent to launch it.'`,
...(docs ? [`echo ' Docs: ${docs}'`] : []),
`echo '${rule}'`
);
@@ -1711,7 +1738,7 @@ function buildMissingCliScript(bin: string, provider: AgentProvider): string {
// ─── IPC: pty lifecycle ─────────────────────────────────────────────────────
/** Spawn options shared by the `pty:spawn` IPC handler and the god-triggered
* ephemeral-worker watcher. */
-type AgentSpawnOptions = SpawnOptions & { hive?: AgentMeta; isolate?: boolean; resume?: boolean; resumeSessionId?: string; provider?: AgentProvider };
+type AgentSpawnOptions = SpawnOptions & { hive?: AgentMeta; isolate?: boolean; resume?: boolean; resumeSessionId?: string; provider?: AgentProvider; noAutoInstall?: boolean };
ipcMain.handle('pty:spawn', async (evt, opts: AgentSpawnOptions) => {
if (!opts || typeof opts.id !== 'string' || typeof opts.cwd !== 'string' || typeof opts.command !== 'string') {
@@ -1743,16 +1770,19 @@ async function spawnAgentCore(opts: AgentSpawnOptions, owner: Electron.WebConten
// just dies with "— process exited (code 1) —" and the user has no idea why.
// Detect the absent binary BEFORE spawning and, in this SAME terminal, print a
// banner + RUN the provider's install command so the user can watch it (and
- // complete any interactive sign-in), then restart the agent. STRICTLY pre-spawn:
- // a non-zero exit from a CLI that DID start never reaches here, so there is no
- // install loop. Providers with no known installer get a manual hint only —
- // nothing arbitrary is ever auto-run. We short-circuit BEFORE worktree/hive/
- // Claude-flag setup: ptyToAgent + worktreePaths stay unset for this id, so when
- // the install PTY exits teardownPty is a harmless no-op (the agent isn't archived
- // and no worktree is torn down) — the user just clicks "restart & continue".
+ // complete any interactive sign-in). On a CLEAN install exit the PTY-exit handler
+ // auto restart-and-continues — it re-runs THIS spawn (with noAutoInstall) so the
+ // freshly-installed CLI launches in the SAME pty/window, no user click. STRICTLY
+ // pre-spawn: a non-zero exit from a CLI that DID start never reaches here, so there
+ // is no install loop; and the relaunch's noAutoInstall guarantees the installer
+ // can't fire twice. Providers with no known installer get a manual hint only (and
+ // are NOT armed for relaunch) — nothing arbitrary is ever auto-run. We short-circuit
+ // BEFORE worktree/hive/Claude-flag setup: ptyToAgent + worktreePaths stay unset for
+ // this id, so when the install PTY exits teardownPty is a harmless no-op (the agent
+ // isn't archived and no worktree is torn down) before the relaunch takes over.
{
const bin = opts.command.trim().split(/\s+/)[0] || opts.command;
- if (bin && !ptyManager.isCommandAvailable(bin)) {
+ if (bin && !opts.noAutoInstall && !ptyManager.isCommandAvailable(bin)) {
const res = ptyManager.spawn(
{
id: opts.id,
@@ -1764,6 +1794,14 @@ async function spawnAgentCore(opts: AgentSpawnOptions, owner: Electron.WebConten
},
owner
);
+ // Arm auto restart-and-continue: when this installer PTY exits cleanly, the
+ // exit handler re-runs the spawn so the just-installed CLI launches in place
+ // (no user click). Only when an installer actually RAN (a provider with no
+ // bundled installer just prints a manual hint and exits 0 — relaunching there
+ // would spawn the still-missing binary and die) and the PTY actually started.
+ if (res.ok && installInfoForProvider(provider).command) {
+ pendingInstallRelaunch.set(opts.id, { opts, owner, bin });
+ }
syncKeepAwake();
return res;
}
diff --git a/src/main/pty.ts b/src/main/pty.ts
index 028d3b56..278652d7 100644
--- a/src/main/pty.ts
+++ b/src/main/pty.ts
@@ -78,7 +78,7 @@ export class PtyManager {
* externally), so the main process can run the SAME lifecycle teardown
* (archive, worktree removal, map cleanup) that the explicit kill() path
* runs. Best-effort — set once by the main process. */
- private exitHandler: ((id: string) => void) | null = null;
+ private exitHandler: ((id: string, exitCode?: number) => void) | null = null;
/** The default/fallback output sink — set to the PRIMARY window. Used only for
* sessions with no recorded owner; owned sessions route to their owner. */
@@ -104,8 +104,10 @@ export class PtyManager {
}
/** Register the natural-exit teardown callback. Invoked from inside node-pty's
- * onExit after the session is cleaned up. */
- setExitHandler(handler: (id: string) => void): void {
+ * onExit after the session is cleaned up. The exit code is forwarded so the
+ * handler can distinguish a clean exit (e.g. a successful first-time CLI
+ * install → auto restart-and-continue) from a crash. */
+ setExitHandler(handler: (id: string, exitCode?: number) => void): void {
this.exitHandler = handler;
}
@@ -309,7 +311,7 @@ export class PtyManager {
this.sessions.delete(opts.id);
// Natural exit must run the same lifecycle teardown as an explicit kill.
// Guarded so a teardown error can never crash node-pty's exit callback.
- try { this.exitHandler?.(opts.id); } catch { /* never throw out of onExit */ }
+ try { this.exitHandler?.(opts.id, exitCode); } catch { /* never throw out of onExit */ }
});
return { ok: true };
diff --git a/src/preload/index.ts b/src/preload/index.ts
index 26cfe625..ea3265c6 100644
--- a/src/preload/index.ts
+++ b/src/preload/index.ts
@@ -456,6 +456,15 @@ const api = {
ipcRenderer.on(channel, listener);
return () => ipcRenderer.removeListener(channel, listener);
},
+ /** Fires when an agent is auto restart-and-continued into this SAME pty after a
+ * first-time engine-CLI install. The terminal should re-arm in place (clear the
+ * "process exited" line + re-enable input) so the relaunched CLI paints clean. */
+ onPtyRelaunch: (id: string, cb: () => void): (() => void) => {
+ const channel = `pty:relaunch:${id}`;
+ const listener = () => cb();
+ ipcRenderer.on(channel, listener);
+ return () => ipcRenderer.removeListener(channel, listener);
+ },
// ─── Dialog ──────────────────────────────────────────────────────────────
chooseFolder: (): Promise<{ ok: true; path: string } | { ok: false; error: string }> =>
diff --git a/src/renderer/src/components/terminalPool.ts b/src/renderer/src/components/terminalPool.ts
index e9b35b56..8bf86f52 100644
--- a/src/renderer/src/components/terminalPool.ts
+++ b/src/renderer/src/components/terminalPool.ts
@@ -90,6 +90,15 @@ export function acquireTerminal(ptyId: string, theme?: ThemeMap, fontSize = 14):
entry.exited = true;
term.writeln(`\r\n\x1b[2m─ process exited (code ${exitCode}${signal ? `, signal ${signal}` : ''}) ─\x1b[0m`);
}));
+ // A first-time engine-CLI install just finished and the agent is auto
+ // restart-and-continuing into THIS same pty (main re-ran the spawn). Re-arm the
+ // terminal in place — clear the latched exit (so keystrokes flow again) and wipe
+ // the install banner + "process exited" line — so the relaunched CLI's TUI paints
+ // onto a clean, typeable grid. Mirrors resetTerminal but works on this closure.
+ entry.unsub.push(window.cth.onPtyRelaunch(ptyId, () => {
+ entry.exited = false;
+ try { term.reset(); } catch { /* not yet open */ }
+ }));
// ── Copy / paste ──────────────────────────────────────────────────────────
// With an accelerated renderer there is no DOM text, so the browser's native