↓ Skip to main content
9 – 17 UHR +49 8031 3508270 LUITPOLDSTR. 9, 83022 ROSENHEIM
DE / EN
Coding agent Continue Dev, Inc. (acquired by Cursor) IDE, Terminal

Continue

Continue is an open-source coding agent for VS Code, JetBrains and terminal. After Cursor acquired it, version 2.0.0 is final and the repo is dormant.

Licence Apache-2.0
First release May 2023 (repository created)
Interfaces VS Code, JetBrains, Terminal (CLI cn)
Platforms macOS, Linux, Windows
Instruction file .continue/rules/

The seven building blocks in Continue

How Continue implements the building blocks of an AI harness

  1. 1

    Agent loop

    The loop runs in VS Code and JetBrains with the Agent, Chat, Edit and Autocomplete features, and in the terminal with cn as an interactive TUI. cn -p runs a task headless and prints the result to stdout or, with --format json, as JSON.

  2. 2

    Tools and execution environment

    Built-in tools such as Read, Search, Edit, Write and Bash can be extended with MCP servers, added in config.yaml or in the CLI via --mcp.

  3. 3

    Context management

    Rules live as Markdown files in .continue/rules/ in the project; the CLI adds rules via --rule. You reference files in the chat with @.

  4. 4

    State and memory

    In the CLI, sessions can be resumed with cn --resume or /resume. Per the documentation, Continue stores development data locally under .continue/dev_data by default.

  5. 5

    Permissions

    In the CLI, every tool is set to allow, ask or exclude: read-only tools are allowed, writing and Bash ask first. Patterns such as Write(**/*.ts) refine this, and persistent rules live in ~/.continue/permissions.yaml.

  6. 6

    Verification and feedback

    In interactive mode you confirm write and Bash calls one by one. Headless, the CLI excludes all ask tools; write permissions must be granted explicitly there via --allow.

  7. 7

    Orchestration

    The current documentation does not describe multi-agent workflows. The plan (--readonly) and auto (--auto) modes override all other permission rules and can be switched with Shift+Tab.

Models and model access

Which models the harness uses and how they are obtained

Models and model access
Models No fixed default model: models live in the local config.yaml, without cloud configuration since version 2.0.0
Access routes Anthropic, OpenAI, Azure, Amazon Bedrock, Google Gemini, Vertex AI, OpenRouter, Mistral, DeepSeek, Ollama, LM Studio, vLLM and other OpenAI-compatible servers, custom OpenAI-compatible endpoints
Free choice of model Yes
Custom endpoint / LLM gateway Yes
Per the documentation, you add a model to config.yaml with provider: openai, apiBase (for example http://localhost:8000/v1) and your own apiKey. For servers without a chat completions endpoint there is the useLegacyCompletionsEndpoint option.
Features
MCP servers Headless / CI IDE integration Resume sessions
Running it with herdr
Supported by herdr No
State detection Not documented
Integration None
The herdr documentation (version 0.9.5, retrieved on 22 November 2026) lists Continue neither among the integrations supported by herdr nor among those supported by the agent itself.
More about herdr →

Enterprise use

Data processing, controls and pricing model according to the vendor

Maintenance status

Per the README, the repository is no longer actively maintained and is read-only; continue.dev announces the acquisition by Cursor. Bug fixes and security updates are therefore no longer to be expected.

Data processing

Since version 2.0.0, per a notice in the extension, configuration is local only, and sign-in and cloud configuration have been removed. Prompts and code go only to the model providers you configure.

Telemetry

With the final 2.0.0 release, per the README, the team removed anonymous telemetry. Per the documentation, Continue writes development data locally to .continue/dev_data; you define additional destinations in the data section of config.yaml.

Identity and teams

With sign-in gone, there is no longer any central user, role or configuration management. Teams distribute config.yaml, rules and permissions.yaml themselves, for example through the repository.

Pricing

Continue is free under Apache-2.0. There is no paid offering anymore: the former pricing page returns a 404 error and hub.continue.dev is unreachable (checked on 22 November 2026).

Strengths
  • Open-source code under Apache-2.0, per the website still freely available as a foundation for further development
  • A shared config.yaml format for VS Code, JetBrains and the CLI
  • Broad provider choice and custom OpenAI-compatible endpoints via apiBase
  • Fine-grained tool permissions in the CLI with patterns and a persistent permissions.yaml
  • Version 2.0.0 without telemetry and without mandatory sign-in
  • Autocomplete, chat and agent in the IDE, per the documentation also offline with local models
Limitations
  • Per the README no longer actively maintained, repository read-only, final version 2.0.0 from June 2026
  • Cloud platform, sign-in and central team management discontinued
  • JetBrains plugin last updated in March 2026; the team recommends the CLI instead
  • Parts of the documentation still refer to sign-in and the hub, which no longer exist
  • No subagents, hooks or sandbox documented
  • Not currently supported by herdr

Use cases

Typical tasks for this harness

Keep using existing VS Code installations until a migration is planned
Autocomplete and chat in the IDE with local models via Ollama or LM Studio
Single tasks in scripts, CI and Git hooks with cn -p
Foundation for your own forks or internal development under Apache-2.0

Continue is an open-source coding agent for VS Code, JetBrains and the terminal that brings freely chosen language models into development through tools, MCP servers and project rules. As an AI harness, Continue is now complete: after Cursor acquired it, the final version 2.0.0 shipped in June 2026. As of 22 November 2026.

What is Continue?

Continue was developed by Continue Dev, Inc.; per the GitHub API, the continuedev/continue repository was created in May 2023, is licensed under Apache-2.0 and has 36,535 stars. Per the VS Code Marketplace, the VS Code extension has around 4.6 million installs (as of 22 November 2026). Continue is available as an extension for VS Code, as a plugin for JetBrains IDEs and as the CLI cn, which per the documentation is installed via an install script or with npm i -g @continuedev/cli. Per the documentation, all three interfaces read the same config.yaml format.

Is Continue still being developed?

No. The continue.dev website states “Continue has joined Cursor”, and the README on GitHub says the repository is no longer actively maintained and is read-only for all users. The final 2.0.0 release of the VS Code extension shipped on 19 June 2026 per GitHub; per the README, it removed anonymous telemetry and authentication. The CLI stands at version 1.5.47 from 18 June 2026 per npm, the JetBrains plugin at 1.0.67 from 27 March 2026 per the JetBrains Marketplace. The last commit on the main branch, from 21 July 2026, only concerns the documentation. The former platform with its hub is no longer reachable.

Continue as an AI harness

Per the documentation, the CLI runs the same agent as the IDE extensions. In VS Code and JetBrains, Continue offers the Agent, Chat, Edit and Autocomplete features; the cn CLI runs as an interactive TUI or headless with cn -p for scripts, CI and Git hooks. Tools such as Read, Edit, Write and Bash can be extended with MCP servers, and project rules live as Markdown files in .continue/rules/.

Per the documentation on tool permissions, the CLI governs rights through three levels: allow, ask and exclude. Read-only tools are allowed by default, writing and Bash ask first; in headless mode these tools are excluded unless you allow them via --allow. The plan (--readonly) and auto (--auto) modes override all other rules. You resume sessions with cn --resume.

Models and model access

Continue is model-agnostic. The documentation covers Anthropic for Claude, OpenAI for GPT, Google for Gemini, Azure, Amazon Bedrock, Vertex AI, OpenRouter, Mistral and DeepSeek, among others, as well as local models via Ollama and LM Studio. Since no updates are released anymore, no new model profiles will be added.

Custom endpoints are supported per the provider documentation: in config.yaml you add a model with provider: openai, the address as apiBase and your own apiKey; examples named include vLLM, LocalAI and LiteLLM. This lets you run Continue in proxy mode through the AI Gateway, whose product page lists a ready-made setup for VS Code with Continue.

Enterprise use: data protection and governance

Version 2.0.0 is data-frugal: per a notice in the extension, configuration is local only, per the README sign-in and telemetry have been removed, and prompts and code go only to the configured model providers. Per the documentation, Continue writes development data locally to .continue/dev_data; this folder belongs in your deletion policies. There is no longer any central user or role management, so teams distribute config.yaml, rules and permissions themselves.

The lack of maintenance weighs more heavily: without bug fixes and security updates, Continue will over time fall short of the state of the art that Art. 32 GDPR requires for the security of processing. You can keep running existing installations but should plan a migration. For new rollouts we recommend a maintained harness such as Cline, Kilo Code or GitHub Copilot.

Running Continue with herdr and Orca

Continue is not currently supported by herdr: the herdr agent list in version 0.9.5 does not name Continue. Orca lists Continue in its own agent overview with automatic setup, without status reporting or hooks.

How does Continue differ from Claude Code and Codex?

Continue is model-agnostic and strongly rooted in the IDE, with autocomplete, chat and agent in VS Code and JetBrains. Claude Code and Codex come from Anthropic and OpenAI, are actively developed and additionally offer subagents and execution in the cloud. The biggest difference is the maintenance status: since the final 2.0.0 release, Continue no longer receives updates. If you want model freedom in the terminal, OpenCode offers it with active maintenance.

Where it fits: developer harness and enterprise harness

Continue is a harness for developers working in the IDE. For business users without a development environment, CompanyGPT is the matching harness, with MCP tools, enterprise knowledge and roles from Entra ID or Keycloak. The AI Gateway provides the shared cost and governance layer; per the product page, its companion measures agent time for Claude Code, Codex, OpenCode, Pi, GitHub Copilot CLI and Gemini CLI, so you capture Continue through proxy mode. Our overview of AI harnesses covers further coding agents and orchestrators.

From developer harness to enterprise harness

Coding harnesses are tools for developers. Business users, cost control and governance need two more layers.

CompanyGPT – the harness for business users

CompanyGPT gives any language model tools via MCP servers, integrations and n8n, company knowledge via companyRAG, roles from Entra ID or Keycloak, and audit logs – operated in your own cloud tenant (Azure, AWS or Google Cloud) or sovereignly on STACKIT.

Discover CompanyGPT
AI Gateway – cost and governance for every agent

The AI Gateway bundles the model access of your tools and coding agents behind an OpenAI-compatible API: sign-in via Entra ID, Keycloak or Cognito, budgets and hard limits per team and user, cost attribution per project and repository via the Companion, and logging of every request.

View the AI Gateway

Frequently Asked Questions

What is Continue?

Continue is an open-source coding agent for VS Code, JetBrains and the terminal, developed by Continue Dev, Inc. under Apache-2.0. It connects freely chosen language models with tools, MCP servers and project rules and offers Agent, Chat, Edit and Autocomplete. Cursor acquired Continue; per the README, version 2.0.0 from June 2026 is the final release.

Is Continue open source and free?

Yes. The source code is licensed under Apache-2.0 and, per the website, remains freely available; the extension and CLI are free. There is no paid offering anymore, and the former cloud platform with sign-in has been discontinued. Costs arise only with the model providers. Per the README, the repository is read-only and no longer actively maintained.

Which models does Continue support?

Continue is model-agnostic. The documentation covers Anthropic, OpenAI, Azure, Amazon Bedrock, Google Gemini, Vertex AI, OpenRouter, Mistral and DeepSeek, among others, as well as local models via Ollama or LM Studio. OpenAI-compatible servers such as vLLM or LiteLLM work as well. Since no updates are released anymore, profiles for future models are missing.

Does Continue work with a custom endpoint or the innFactory AI Gateway?

Yes. In config.yaml you add a model with provider openai, the gateway address as apiBase and your key. This runs Continue in proxy mode through the AI Gateway, whose product page lists a ready-made setup for VS Code with Continue. The gateway companion's agent time measurement does not cover Continue.

Can Continue be used in a GDPR-compliant way?

Per the README, version 2.0.0 no longer sends telemetry, configuration is local and prompts go only to the configured providers. Their contracts and your record of processing activities remain decisive. The lack of maintenance weighs more heavily: without security updates, Continue will over time fall short of the state of the art that Art. 32 GDPR requires for the security of processing.

How does Continue differ from Claude Code and Codex?

Continue is model-agnostic and strongly rooted in the IDE, with autocomplete, chat and agent in VS Code and JetBrains. Claude Code and Codex come from Anthropic and OpenAI, are actively developed and additionally offer subagents and cloud execution. The biggest difference is the maintenance status: since the final 2.0.0 release, Continue no longer receives updates.

Introduce coding agents safely in your company

We support selection, model access in the EU, permission design and cost control.