Continue.dev

Continue.dev's status changed materially in 2026. The homepage fetched on September 2 says Continue has been acquired by Cursor and that its open-source codebase remains available as a foundation for others. The old pricing route returned a GitHub Pages 404. Builders should therefore evaluate Continue as open-source software with an uncertain service lifecycle, not assume the former hosted product or paid plans remain purchasable.

Before the acquisition, Continue provided chat, autocomplete, inline editing, agent workflows, reusable rules, documentation context, model choice, and Model Context Protocol tools in VS Code and JetBrains. Its Apache 2.0 codebase appealed to platform teams that wanted to select OpenAI, Anthropic, Gemini, or local models instead of accepting one vendor's fixed stack. Teams could assign a low-latency model to completion, a stronger model to chat, and local inference through Ollama or LM Studio when source code could not leave a controlled environment. Those architectural options remain useful, but users must verify which repositories, extensions, releases, and cloud dependencies Cursor will maintain.

There is no defensible current subscription price to publish. Repository evidence previously listed Starter at $3 per million tokens, Team at $20 per seat monthly, and Company on custom terms, but those are historical figures because the live pricing page is gone. A self-managed deployment still incurs model API charges, local GPU capacity, embeddings, security review, extension maintenance, and staff time.

The main advantage is control: teams can inspect the implementation, customize prompts and tools, and avoid one model provider. The main risk is lifecycle uncertainty after an acquisition. Compare maintained options including Aider, Cursor, GitHub Copilot Agents, and Windsurf; the AI coding agents comparison covers broader selection criteria.

Existing users should inventory dependencies on Continue-hosted services, pin a known-good extension, mirror required repositories, test provider authentication, and document an exit path. New pilots should use a disposable repository and confirm that completion, indexing, rules, and MCP connections work without retired services. Run 20 representative tasks and measure success, latency, model cost, false edits, and maintenance time. Continue remains relevant open-source code, but the acquisition makes manual verification essential before an organization standardizes it. Security review should include extension permissions, telemetry, data retention, model-provider terms, dependency provenance, and the handling of repository secrets. Platform owners should also assign responsibility for patching the fork and define how quickly developers can migrate if editor compatibility or hosted endpoints stop working. Those operational details determine whether the apparent flexibility is sustainable.

Procurement should not copy historical hosted-plan prices into a 2026 budget. Confirm extension availability, release ownership, security-patch responsibility, license obligations, and remaining hosted dependencies from maintained repositories and the acquisition notice. A fork may avoid subscription fees while adding engineering and incident-response costs.