Apps · IDEs & editors

Continue

Open-source autopilot for VS Code and JetBrains.

Tokens / 7d

88B

Routing directive

auto:coding

On our metal

31%

Licence

Open source

Continue separates autocomplete, chat and edit roles. Autocomplete is latency-bound and runs on small fill-in-the-middle models; chat escalates to a frontier model only when asked.

→ continue.dev → All apps → Integration docs

How it connects

config.json declares one provider block per role.
Fill-in-the-middle models served from our own fleet for sub-100ms TTFT.
Team keys with per-seat budgets on the Team plan.

Traffic pattern

Extremely high request count, tiny payloads. Time-to-first-token dominates perceived quality.

qwen2.5-coder-32bcodestral-2508claude-sonnet-4.5

Point it at the gateway

{
  "models": [{ "title": "chat", "provider": "openai",
    "apiBase": "https://api.llmcloud.ai/v1", "model": "auto:coding" }],
  "tabAutocompleteModel": { "provider": "openai",
    "apiBase": "https://api.llmcloud.ai/v1", "model": "codestral-2508" }
}