Continue

Continue is an open-source AI coding assistant for VS Code and JetBrains. Its architecture routes chat, tab completion, and inline edits directly to any self-hosted model or commercial API through local configuration. Developers maintain full control over where prompts, code context, and weights reside.

Why Continue instead of Cursor AI

Cursor AI delivers assistance by distributing a customized fork of Visual Studio Code. While that fork provides tight editor integration, it introduces operational trade-offs for infrastructure and security teams: developers must abandon standard editor distributions, update channels, and security-reviewed binaries for a vendor-managed application. Cursor also routes context and agent actions through its proprietary cloud backend by default. In contrast, Continue operates as an extension inside standard VS Code and JetBrains installations, preserving existing workstation management and extension policies.

The operational divide centers on model hosting and network boundaries. Continue delegates inference to endpoints declared in a config.yaml file, supporting local runtimes like Ollama and vLLM alongside internal enterprise gateways. For organizations with strict data residency mandates or air-gapped environments, Continue keeps source files and completions entirely inside the corporate perimeter. While Cursor offers a smoother out-of-the-box experience, Continue gives DevOps teams auditable model routing, zero mandatory telemetry, and freedom from proprietary cloud lock-in.

What it does well

• Multi-model routing: Directs chat, autocomplete, and edits to separate backends, letting teams pair fast local models for completions with larger models for refactoring.
• Declarative configuration: Stores model selections, rules, and tool settings in version-controlled .continue/config.yaml files checked into repositories.
• Air-gapped inference: Integrates natively with Ollama and vLLM, enabling code generation on local GPUs without sending data across external networks.
• Model Context Protocol support: Connects agent workflows to external context through MCP servers, providing controlled access to issue trackers and internal documentation.
• Granular context providers: Indexes local codebases with configurable embeddings and supports targeted references like @docs, @diff, and @terminal in chat sessions.