Move your team from Cline to Continue.
Each person gets a short interview, then a personalized step-by-step guide showing exactly where their saved work, prompts, and projects land in Continue — and the few things that need a manual step, walked through. You send one invite; nobody becomes the help desk, and nothing in Cline is changed or cancelled. Want to see where you’re paying for both first? Run the free audit.
Both tools support headless and CLI-driven automation that runs the coding agent outside the IDE, suitable for CI/CD pipelines and scripted workflows. In Cline, the CLI is installed via npm (npm i -g cline) and the --auto-approve flag enables fully headless operation without per-call confirmation dialogs. In Continue, the CLI is cn (installed via curl/PowerShell/npm, requires Node.js 20+) and headless mode is triggered with cn -p 'prompt', outputting only the final response for use in Unix pipelines. To migrate: replace cline --auto-approve 'task' with cn -p 'task' in CI scripts; use --allow flags in Continue to control which tool categories are auto-approved instead of Cline's YOLO/auto-approve modes. Both CLIs share config with their respective IDE extensions, so API keys and rule files already in place are reused automatically.
- Warning: Cline's terminal-execution capability also covers interactive in-IDE shell command execution with granular auto-approve controls (safe commands only, all commands, or YOLO full auto-approve). Continue's cli-headless-agents is a separate CLI binary (cn) for running the agent headlessly — it does not replace terminal command execution within the IDE. In-IDE shell execution in Continue is handled by Agent Mode through the built-in terminal tool, not the cn CLI.
- Warning: Continue's CLI (cn) requires Node.js 20+. The former CONTINUE_API_KEY hosted-authentication path for CI is defunct: the final open-source release (2.0.0, June 19, 2026) pulled authentication out of the codebase, so cn runs purely locally with provider API keys, same as Cline's ~/.cline config approach.
- Warning: Continue's Mission Control cloud agents and hosted PR-check workflows no longer exist — the hosted platform was shut down after Cursor acquired Continue in June 2026, and the continuedev/continue repository is read-only and no longer actively maintained. The open-source cn CLI still works locally, but this migration lane targets an unmaintained product; teams should weigh that before adopting Continue as a destination.
Both tools let users inject targeted context into the chat via @-prefixed references to files, folders, problems, terminal output, and URLs, so the agent has precise context without manual copy-paste. In Cline, typing @/path/to/file injects file content, @/path/to/folder/ injects folder structure and contents, @problems surfaces workspace diagnostics, and URL mentions fetch and inject web content. Files can also be dragged into chat. In Continue, typing @ opens a picker with context types: @File, @Code (symbol), @Git Diff, @Current File, @Terminal, @Open (all open files), @Clipboard, @Tree, @Problems, @Operating System, and @HTTP (custom endpoint). To migrate workflows: Cline's @file → Continue's @File; Cline's @folder → Continue's @Tree or @Open; Cline's @problems → Continue's @Problems; Cline's URL mention → @HTTP (or an MCP server for richer doc retrieval).
- Warning: Several Continue context providers that users may expect (@Codebase, @Docs, @Folder, @URL, @Web, @Search, @Database, @Jira, @GitHub Issues, @Commits) are deprecated in Continue's current version in favor of MCP server equivalents. Teams migrating from Cline who relied on @URL-style context injection must set up MCP servers (e.g. Context7, DeepWiki) to replicate this in Continue.
- Warning: Cline supports drag-and-drop of files (text, images, PDFs, CSVs, Excel) into chat as a quick context injection method. Continue does not document an equivalent drag-and-drop interface for non-code files — binary file injection must be handled via @File or custom MCP tooling.
Both tools support the Model Context Protocol standard for extending the agent with external data sources, APIs, and custom tooling. Cline stores server configs in ~/.cline/mcp.json (CLI) or extension settings and offers a visual MCP Marketplace tab for one-click installs. Continue stores server configs locally in .continue/mcpServers/ (YAML or JSON) using the mcpServers property — the former Mission Control hub-config path is defunct since the hosted platform shut down after Cursor's June 2026 acquisition of Continue. To migrate MCP server configs from Cline to Continue: export your Cline MCP config (JSON format from ~/.cline/mcp.json), then place that JSON file in .continue/mcpServers/ — Continue explicitly documents accepting Cline-exported MCP configs. Secrets previously hardcoded in Cline's config should be moved to Continue's ${{ secrets.KEY_NAME }} injection syntax for safety. Both support stdio, SSE, and HTTP transports. Re-test each MCP tool in Continue's Agent Mode (MCP tools are not available in Chat or Plan Mode).
- Warning: Cline's MCP Marketplace provides a curated one-click install hub with hundreds of pre-configured servers. Continue has no equivalent marketplace UI — all MCP servers must be configured manually via local YAML/JSON files (Mission Control is gone, and the continuedev/continue repo is read-only/unmaintained after the June 2026 Cursor acquisition). Discovery of new MCP servers requires visiting the community ecosystem directly.
- Warning: In Continue, MCP tools are only available in Agent Mode. Cline exposes MCP tools in Act mode but the marketplace approval UI persists across modes. If a team relies on agent-gated MCP call approvals, Continue's per-call 'Continue/Cancel' approval still applies in Agent Mode.
- Warning: Continue's MCP support notes compatibility with configs exported from Claude, Cursor, and Cline in JSON format, but YAML format (the native Continue format for mcpServers/) requires the server definitions to be translated manually.
Both tools support bring-your-own-key (BYOK) for all major cloud model providers and local runtimes with no locked-in model. In Cline, API keys and provider selection are set in the extension settings UI or the ~/.cline config file; the Cline Provider path also offers managed billing with OAuth. In Continue, configuration is done in local config.yaml (the hosted Mission Control path is defunct — the platform shut down after Cursor's June 2026 acquisition of Continue, which also left the open-source project read-only/unmaintained) using a models array where each entry includes provider, model, apiKey (or ${{ secrets.KEY_NAME }}), and role assignments. To migrate: identify which cloud providers you use in Cline (Anthropic, OpenAI, Gemini, Bedrock, OpenRouter, DeepSeek, etc.), then add the equivalent entries in Continue's config.yaml under the models array. Assign roles (chat, edit, autocomplete) to ensure each use case has a model. Local model runtimes (Ollama, LM Studio) are supported identically on both sides with the same base URL convention.
- Warning: Cline offers a managed 'Cline Provider' billing path with OAuth and free model access. Continue has no equivalent managed billing — all model costs must be paid directly to the provider using your own API key. Teams that relied on Cline's managed billing tier will need to set up direct provider accounts.
- Warning: Cline supports per-mode model selection (different model for Plan vs Act). Continue assigns models by functional role (chat, edit, autocomplete, embeddings) rather than by plan-vs-agent phase. There is no direct equivalent for using a different model in Plan Mode vs Agent Mode without manual config editing.
- Warning: Continue supports significantly more providers in config (SambaNova, WatsonX, SageMaker, Nebius, Cloudflare, NVIDIA, Together AI, DeepInfra, etc.) that are not listed as Cline providers. Teams with specialized enterprise model hosting may find better options available on the Continue side.
Both tools support persistent markdown instruction files that guide the AI's behavior across sessions without repeating preferences manually. In Cline, workspace rules live in a .clinerules/ directory at the project root (checked into version control); global rules are at Documents/Cline/Rules. In Continue, workspace rules go in .continue/rules/ (project-scoped) and global rules in ~/.continue/rules/. Both use markdown format. To migrate from Cline to Continue: copy your .clinerules/ markdown files into .continue/rules/ at the project root. Review Cline's YAML frontmatter glob conditions (used for conditional rule activation) and replicate them in Continue's equivalent globs or regex fields in the rule YAML front matter. Cline also recognizes .cursorrules and AGENTS.md; Continue does not natively consume those files, so explicit migration to .continue/rules/ is required.
- Warning: Cline uses YAML frontmatter with glob patterns inside .md files to activate rules conditionally (e.g. frontend rules only when editing React components). Continue uses a similar mechanism with globs, regex, and alwaysApply fields — but the frontmatter field names differ, so conditional rules need to be re-written rather than copied verbatim.
- Warning: Cline rules apply in both Plan and Act modes. Continue rules apply to Agent, Chat, and Edit modes but NOT to autocomplete. If any rules are intended to influence autocomplete behavior, they must instead be configured at the model/autocomplete level in Continue's config.yaml.
- Warning: Cline's /newrule slash command scaffolds new rule files interactively in the IDE. Continue has no equivalent scaffolding command but Agent Mode can auto-create rules via the create_rule_block tool.
Both tools split the agent workflow into a read-only planning phase and a full-execution phase. In Cline, Plan mode is switched via the Plan/Act toggle in the chat UI and prevents file writes or command execution; Act mode enables all tools, with conversation context carried over on switch. In Continue, the same split is available in the mode selector dropdown below the chat input: Plan Mode (read-only tools for safe exploration) vs Agent Mode (all tools, including file writes and terminal commands); Cmd/Ctrl+. cycles between modes. When switching from Cline to Continue: select 'Plan Mode' in the Continue dropdown to replicate Cline's Plan phase; switch to 'Agent Mode' to replicate Act phase. Continue also offers a plain 'Chat' mode with no tools, which has no direct Cline equivalent. The model used for each phase is set in Continue's config.yaml under separate model role assignments rather than per-mode selectors as in Cline (Cline exposes a 'Use different models for Plan and Act' setting).
- Warning: Cline allows per-mode model selection (e.g. a stronger reasoner for Plan, a faster model for Act) via the extension UI. Continue assigns models by role (chat, edit, autocomplete) in config.yaml, not by plan vs agent phase — you cannot trivially assign a different model to Plan Mode vs Agent Mode without manual config edits.
- Warning: In Continue, MCP tool extensions are available in Agent Mode only; Plan Mode uses only built-in read tools. Cline's Plan mode similarly restricts to reads, but MCP tool availability in Plan is not separately gated by the marketplace UI.