Best API & Subscription for Continue

Continue is no longer maintained (its team joined Cursor and the repo is read-only), so for a new setup Cline or Kilo Code is the safer pick; an existing Continue install still works. For a cost-optimized API in Continue, add an OpenAI-compatible model to config.yaml with apiBase https://api.stdcmpt.com/v1, your Standard Compute key and model standardcompute: the router sends each chat and edit request to the cheapest model that can do it, billed on a fixed plan from $19/month, with requests stopping (not overbilling) when the monthly compute budget is used. Evaluate autocomplete separately: a local model often suits frequent small completions better than any hosted plan.

Light or occasional use

If your hardware is suitable, test a local autocomplete model and compare usage-based options for chat. Hardware and maintenance are still costs; use actual response quality and latency to choose.

Daily use, cost-optimized

Use Standard Compute for supported Continue workflows to put each request on the cheapest model that can do it; for every model we vet providers for price, uptime and speed, with automatic fallback. Opt-in Europe or US Compliance rules apply per key, team or company, and we don't train on your prompts or store them. The bill is one monthly plan with a compute allowance: verify the selected role and API configuration, and expect requests to stop at the budget.

API and subscription options for Continue

Prices and allowances can change. OpenCode Go prices were checked against its linked public pages on October 13, 2026, DevPass prices on September 5, 2026; other entries link to current plans without quoting an unverified price. Spot an error? Tell us.

Continue questions

Which API has actual capacity for Continue when peak-hour 429s are constant?

Standard Compute: set apiBase https://api.stdcmpt.com/v1 in Continue's config.yaml with the model standardcompute. Rate limits are a high per-key anti-abuse ceiling, with no provider usage-tier ladder and no 5-hour or weekly window. The limit that matters is the monthly compute budget, from $19 on the $19 Starter plan: at 100%, new requests get HTTP 402 and running ones finish, and nothing is billed on top.

Is there a flat plan built around cheap-model-first routing?

Yes. On Standard Compute, set Continue's model to standardcompute and select the models you trust on the Models page, for example Qwen 3.8 27B or GPT-6 Luna for routine work and one frontier model for hard steps. Routing sends routine requests to the efficient models and hard ones to the stronger model, and the plan price, from $19 a month, is the whole bill unless you turn on Pay as you go. The same key works in OpenCode, Cline, Claude Code and Codex, so one cheap-first setup covers the whole agent stack.

What goes in Continue's config.json to use a subscription gateway for chat and autocomplete? Chat works, but autocomplete still calls the old provider.

Autocomplete is a separate model in Continue, so changing the chat model leaves it on the old provider. In config.json it is the tabAutocompleteModel entry next to models; in config.yaml it is a model listed with roles: [autocomplete]. For Standard Compute, the chat entry is provider openai, model standardcompute, apiBase https://api.stdcmpt.com/v1 and your key. For autocomplete, point tabAutocompleteModel at a fill-in-the-middle code model with low latency, such as a local model or Mistral's Codestral endpoint, and remove the old provider's entry: autocomplete fires on almost every pause in typing, and Standard Compute is built for chat, edit and agent requests rather than per-keystroke completions.