The Pulse
Amp Makes Its Coding Agent Free for BYOK Users
Amp says its new free Hobby tier lets developers use their own model subscriptions or API keys without a monthly Amp plan or BYOK token fees, while remote orb computing remains paid.
AI.info Team ·
Amp is removing the subscription barrier for developers who bring their own model access. The coding-agent company announced on September 13, 2026, that users can run Amp with ChatGPT subscriptions, other model subscriptions, or their own API keys without paying Amp a monthly plan or token fees.
“Amp is now free to use when you bring your own compute and model subscriptions/keys. No more limits or fees for BYOK.”
Quinn Slack, CEO of Amp, in a post on X
The offer is not a blanket promise of free computing. Amp will continue charging for its remote “orbs,” which run agents independently and in parallel, while developers can use “runners” on their own machines at no charge. The distinction puts model access and compute on separate bills: users can avoid Amp’s inference charges, but remote execution still costs money unless they supply the hardware.
Amp’s Hobby tier removes the monthly gate
The new Hobby tier costs nothing and includes Amp’s full product feature set, according to the company. Users can pay as they go for orbs, run agents on their own computers, use tokens from any supported source, connect ChatGPT and other subscriptions, or configure their own provider keys.
Amp says Hobby users face no Amp token fees or token limits, with both public and private repositories supported. Developers who prefer to buy inference through Amp can continue doing so without a markup, but the new structure makes that option voluntary rather than required for using the product.
The change also separates Amp from the limits imposed by an individual model subscription. A developer can connect more than one provider or route requests through a company’s preferred AI gateway instead of depending on the allowance attached to a single ChatGPT plan.
Free model access, paid remote machines
Amp’s orbs remain the company’s main paid resource. They are remote computers where agents can work independently and in parallel, while runners provide a free alternative for users willing to run the workloads on their own hardware.
That pricing boundary matters for teams using agents on longer tasks. BYOK removes Amp’s charge for model tokens, but it does not make remote execution free. Users who need persistent or parallel remote environments must still pay for orb usage, whereas a local runner shifts the infrastructure cost to the developer or organization.
Amp also retains its paid Megawatt and Gigawatt plans. The company says Megawatt members receive an average 60% discount on Amp usage and Gigawatt members receive an average 65% discount. Those members keep their discounted orb and model allocations while gaining access to additional token sources.
Teams can add users without new seats
Amp is pairing the free Hobby tier with a Teams tier that carries no additional charge. Teams can place members on different plans, including Hobby, Megawatt, or Gigawatt, while sharing threads, portals, plugins, skills, billing, pooled credits, and administrative controls.
Every member can use a ChatGPT subscription or a company-selected AI gateway through BYOK without a monthly Amp fee. Power users can pay for a higher tier and discounted orb access, while other members remain on Hobby and use pooled workspace credits for orb consumption on a pay-as-you-go basis.
Amp says workspaces also receive SAML or OIDC single sign-on at no extra charge when at least one member is on a paid tier. The company is extending its zero- and minimal-data-retention policy to all users and workspaces, rather than contractually guaranteeing it only to Enterprise customers.
More providers arrive first for paid members
Megawatt and Gigawatt members receive early access to additional BYOK connections. Amp lists OpenRouter, Amazon Bedrock, Google Cloud Agent Platform, Azure Foundry, Vercel AI Gateway, Cloudflare AI Gateway, Ollama Cloud, OpenCode Go, and custom endpoint URLs.
The new connections require users to enable Amp’s “More AI Routers & Subscriptions” experimental feature and configure providers in personal or workspace settings. Amp says it plans to make the options available to everyone later.
Enterprise remains the exception to the no-limit BYOK policy. Amp’s announcement says token fees and limits do not apply to users except those on the Enterprise tier, whose offering instead includes pooled credits, directory synchronization, spend controls, audit logs, network and MCP controls, regional endpoints, higher quotas, and custom commercial terms.
The new economics of using Amp
Amp’s offer makes its coding agent free when developers supply the model access and, where necessary, the compute. The company earns from remote execution, paid subscriptions with discounted orb time, and customers who continue buying inference through Amp.
For individual developers, the practical choice is between running agents locally at no Amp charge or paying for Amp’s managed machines. For teams, the announcement lowers the cost of adding occasional users, but shared remote workloads and higher execution quotas remain tied to paid resources.