Clark
Clark is a coding agent you can run either in the cloud, on tasks that don't need your local setup, or directly on your own machine when it does — instead of forcing every task into one mode.
🔗 Visit Clark
Description
Most AI coding agents pick one deployment model: either they run entirely in the cloud (fast to spin up, but disconnected from your local environment and files) or entirely on your machine (full access, but tied to that one computer). Clark offers both, letting a team route a task to whichever mode fits — cloud for parallel or background work, local when a task needs your actual dev environment.
Clark supports team collaboration with shared credits, bring-your-own-key (BYOK) for using your own model API keys, organization-level billing controls, and specialized "Coder" plans built around its Clark Code workflows. Pricing is tiered (Starter, Pro, Scale) with region-based rates set dynamically through Stripe at checkout, plus seat-based team plans and per-token API pricing — though specific numbers aren't published upfront.
💬 Our review
The short version: Clark is for a dev team that wants the flexibility of both cloud and local coding agent execution under one billing relationship, rather than juggling two separate tools for the two modes.
The hybrid cloud/local model addresses a real limitation in most single-mode coding agents — cloud-only agents can't touch your actual dev environment, and local-only agents can't easily parallelize background work across a team. Offering BYOK is a meaningful cost-control option for teams that already have API relationships with model providers and don't want to pay a markup. Organization billing controls suggest this is built with team procurement in mind, not just individual developers. The main friction is pricing opacity — region-based dynamic Stripe pricing with no published numbers means you can't compare cost against Cursor, Copilot or Codex until you actually go through checkout, which is an unusual amount of friction for a developer tool evaluation.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Starter/Pro/Scale, prix dynamique par région via Stripe ; plans équipe au siège ; API au token
Pros
Combine exécution cloud et locale
Bring-your-own-key pour maîtriser les coûts
Contrôles de facturation par organisation
Crédits partagés en équipe
Cons
Tarifs non publiés à l'avance (dynamiques par région)
Difficile de comparer le coût avant de passer au paiement
Moins établi que Cursor ou Copilot