Continue
Continue is the open-source AI code assistant for VS Code and JetBrains, distributed under Apache 2.0. The Continue IDE extensions and the Continue CLI federate to any LLM provider — Anthropic, OpenAI, Mistral, OpenRouter, Ollama, and a Continue-managed proxy — and load their configuration from Continue Hub. Continue Hub (api.continue.dev) is the registry and IDE API that serves assistants, blocks, models, rules, prompts, docs, and MCP servers to the extensions, along with free-trial status, organization policy, secrets sync, and the Stripe checkout URL for the Models Add-On. Continue has also pivoted into "continuous AI" — source- controlled checks that live as markdown files under .continue/checks/ and run in CI as GitHub status checks.
APIs
Continue Hub IDE API
REST API exposed by api.continue.dev that the Continue VS Code and JetBrains extensions and the Continue CLI use to list and refresh assistants, list organizations, fetch organi...
Continue IDE Extensions
Open-source IDE plugins shipping for VS Code and JetBrains. Provide chat, edit, apply, autocomplete, and agent modes. Bring your own LLM (Anthropic, OpenAI, Mistral, OpenRouter,...
Continue CLI
Open-source command-line interface powering Continue's source-controlled AI checks. Checks are markdown files in .continue/checks/ that run as full AI agents against pull reques...
Continue Hub
Hosted registry where teams publish and share assistants, blocks (models, rules, prompts, docs, MCP servers, context providers), and policy. Powers the Continue Hub IDE API and ...
Continue Mission Control
Observability surface for tracking check outcomes, agent runs, and adoption metrics across an organization's repositories. Part of the Team and Company tiers.
Collections
Continue Hub IDE API
OPENPricing Plans
Rate Limits
FinOps
Continue Dev Finops
FINOPSFeatures
Use Cases
Use any frontier or local model directly inside VS Code and JetBrains without lock-in to one vendor.
Define checks as markdown in .continue/checks/, have AI enforce them on every PR, see results as GitHub status checks.
Publish assistants on Continue Hub so every engineer pulls the same models, rules, prompts, docs, and MCP servers.
Federate to a self-hosted Ollama, an OpenRouter account, or your own Anthropic/OpenAI key to keep AI spend on your existing provider.
Route model traffic through an on-prem Continue proxy so prompts and secrets never leave the corporate network.
Run code completion against a local model for zero-egress, zero-cost inference.
Compose MCP servers into a Continue assistant to give the agent access to internal tools without writing extension code.
Apply Continue's "standards as checks" model to legacy codebases — codify standards in markdown, then have AI agents drag the codebase toward them.
Integrations
Use Claude Opus, Sonnet, and Haiku as the chat/edit/apply/autocomplete model inside Continue. Anthropic-continue-hub publishes ready-to-use blocks.
Use GPT-class models as Continue model providers via openai-continue-hub.
mistral-continue-hub publishes Mistral models as Continue blocks.
openrouter-continue-hub federates to any OpenRouter-hosted model.
ollama-continue-hub points Continue at a local Ollama server for zero-egress inference.
bedrock-continue-hub uses Bedrock-hosted models, including Claude on Bedrock.
google-continue-hub federates to Gemini and Vertex AI models.
Per-provider Continue Hub repositories ship blocks for each.
Continue checks run in CI and report status checks back to the GitHub PR. The check-cli, checks-cli, and suggestions-cli automate these flows.
Starter tier and above can connect Continue agents to Slack, Sentry, and Snyk.
docker-continue-hub publishes Docker-authored blocks for Continue.
Continue assistants compose MCP servers as tool sources; Continue maintains a fork of the official MCP TypeScript SDK.
Solutions
Codify coding standards as markdown checks; AI agents enforce them in CI on every pull request.
Drop-in replacement for proprietary AI coding assistants without surrendering model choice or telemetry control.
Centralize team assistants on Continue Hub instead of per-engineer config sprawl.