Openbase
Lets you manage your Claude Code, Cursor or Codex coding agents by talking to them from your phone, so you can dispatch and approve tasks without sitting at your desk.
🔗 Visit Openbase
Description
AI coding agents can now write code, open pull requests and run tests mostly unsupervised, but you usually still need to sit at a keyboard to steer them, approve risky actions, or check what they're doing. Openbase removes that constraint by turning the whole interaction into a voice conversation you can have from your phone, on a walk, or away from your desk.
Openbase is a voice-controlled interface that works across Claude Code, Codex, Cursor and Gemini CLI, letting you dispatch tasks, steer agents mid-run, and approve or reject changes by speaking rather than typing. It ships as a macOS app, an iOS app and a web console, with a diff-review interface for checking code changes, an approval workflow gate for sensitive actions, and live transcription of your spoken commands. It's in open beta with active development and has open-source components on GitHub, but no published pricing yet — the current focus is a waitlist rather than paid plans.
💬 Our review
The short version: if you're already running multiple AI coding agents in parallel and getting tied to your desk just to approve their next move, Openbase's voice control is a genuinely useful unlock — talk to your agents like you'd talk to a junior developer over the phone.
It's a new category rather than a direct swap for any single existing tool — it sits on top of Claude Code, Cursor or Codex rather than replacing them, closest in spirit to Heard (which gives agent output a voice) but going further by adding voice input and control, not just voice output. The catch is it's early-stage: no public pricing, beta status, and success depends heavily on how reliably it interprets spoken commands for something as precise as code changes — voice-driven approval of code diffs is inherently riskier than reading them on screen. Worth trying if you already run agents unsupervised and want mobility; not for anyone uncomfortable approving code changes without reading every line first.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Pas de page tarifaire publique ; produit en bêta ouverte, liste d'attente active
Pros
Contrôle vocal complet : dispatch, pilotage en cours d'exécution, approbation
Fonctionne avec plusieurs agents (Claude Code, Codex, Cursor, Gemini CLI)
App macOS + iOS + console web pour piloter depuis n'importe où
Composants open-source disponibles sur GitHub
Cons
Aucune tarification publique, produit en bêta
Valider des changements de code à la voix est plus risqué qu'une lecture à l'écran
Dépend de la fiabilité de la reconnaissance vocale pour des commandes précises
Catégorie encore naissante, peu de recul utilisateur