Continue is an open-source coding agent available
through its IDE integrations and command-line
interface. Its current workflow includes Chat,
Plan and Agent modes, allowing developers to
move from conversation and read-only planning
to tools that can edit files and run commands.
Continue is an open-source AI coding platform
built around customizable models, rules and
tools. Developers can use its coding agent
through supported IDE integrations or its
terminal-based CLI.
The platform is designed for more than simple
autocomplete. Agent mode can use tools to inspect
a codebase, create files, edit existing files,
run terminal commands and interact with external
services.
Who Is Continue For?
💻 Software developers
🧑💻 Freelancers
🎓 Programming students
🚀 Startup teams
🛠 Open-source developers
🏢 Engineering teams
⚡ Developers building custom AI workflows
Key Features
Continue AI Features
Continue combines conversational AI with
development tools that allow models to
understand and modify software projects.
🤖
Agent Mode
Give the model tools to perform multi-step
coding tasks and modify your project.
📋
Plan Mode
Explore a project using read-only tools
before making implementation changes.
💬
Chat Mode
Use pure conversation without tool access
for questions, explanations and discussion.
📂
Read Files
Inspect project files and provide relevant
code context to the selected model.
✏️
Edit Files
Agent mode can make changes to existing
files within the project.
➕
Create Files
Generate new files as part of a
multi-step development task.
💻
Terminal Commands
Run commands from the workspace root
when the agent has permission.
🔎
Code Search
Use grep, glob and other tools to
find relevant project code.
🗺️
Repository Map
Give the model structural information
about the repository.
🌐
Web Search
Use web-related tools when supported
to retrieve useful development context.
🔌
MCP
Connect external tools and services
through Model Context Protocol.
⚙️
Custom Rules
Configure rules and system instructions
for coding workflows.
Three Modes
Chat, Plan & Agent Modes
Continue separates conversational assistance,
safe exploration and active coding into
different modes.
💬
Chat Mode
Conversation without coding tools.
Ask programming questions
Explain code
Discuss architecture
Generate ideas
Review concepts
📋
Plan Mode
Read-only tools for safe exploration
and planning.
Read files
Search code
View repository structure
Inspect diffs
Create implementation plans
⚡
Agent Mode
Tools for actively making changes
to the codebase.
Create files
Edit files
Run terminal commands
Use external tools
Complete multi-step tasks
Agent Mode
How Continue Agent Mode Works
1
Send Request
Describe what you want the
coding agent to accomplish.
2
Choose Tools
The model can select appropriate
tools for the requested task.
3
Ask Permission
Continue can request approval
before tool execution.
4
Return Result
Tool results return to the model
for the next reasoning step.
Continue's default tool policy asks for
permission. Tool policies can also be
configured as automatic or excluded,
depending on the workflow.
Codebase Tools
Understand Your Existing Project
📖
Read File
Inspect the contents of a
specific project file.
📄
Current File
Provide the currently open
editor file as context.
📁
List Directory
Explore folders and understand
project structure.
🔍
Glob Search
Find files matching project
patterns.
🔎
Grep Search
Search source code for symbols,
strings and patterns.
🗺️
Repo Map
Understand important structural
information about the repository.
📊
View Diff
Inspect differences between
code states.
🌐
Fetch URL
Retrieve useful web content
for supported workflows.
🧠
Codebase Tool
Use codebase-aware tooling to
retrieve relevant context.
Multi-Step Development
Build Features With Agent Mode
Example Task
Add authentication to this application.
1. Inspect the existing project.
2. Find the user model.
3. Identify the current routing system.
4. Create login and registration screens.
5. Add authentication logic.
6. Add validation.
7. Write tests.
8. Run the tests.
9. Fix failures.
10. Explain every changed file.
What Continue Can Do
📂 Explore the repository
🔎 Search relevant code
✏️ Edit existing files
➕ Create new files
💻 Run commands
🧪 Run tests
🔁 Iterate after errors
📊 Review diffs
The selected model must support the required
tools. Continue's documentation notes that
tool availability varies by model and provider.
MCP
Extend Continue With External Tools
Model Context Protocol can add external
capabilities to Continue's agent workflow.
🔌
External Tools
Connect services that expose
compatible MCP tools.
🗄️
Databases
Connect database-related MCP
servers for suitable workflows.
🌐
Web Services
Provide agents with access to
additional web-based capabilities.
📚
Documentation
Connect external sources of
documentation and knowledge.
🛠️
Custom MCP Servers
Build custom integrations for
your own development workflow.
⚙️
Tool Policies
Choose whether tools ask,
run automatically or are excluded.
Rules & Customization
Customize How Continue Codes
Rules
Continue allows developers to define rules
that can guide the coding agent's behavior,
style and project-specific expectations.
Code style requirements
Architecture rules
Testing requirements
Documentation conventions
Project-specific instructions
System Messages
Model-level configuration can customize
system messages for Chat, Agent and Plan
modes.
models:
- name: Coding Agent
provider: openai
model: your-model
chatOptions:
baseAgentSystemMessage:
"You are a systematic coding agent."
CLI
Continue From The Terminal
Continue CLI, available as cn,
brings the coding agent into the terminal.
It supports interactive and headless workflows.
1
Install
Install the Continue CLI using
the supported installation method.
2
Open Project
Move into the project directory
from your terminal.
3
Start Agent
Run cn for an
interactive coding session.
4
Automate
Use headless prompts for scripts,
CI/CD and git hooks.
Interactive CLI
TUI Mode
The interactive terminal interface lets
developers communicate with the coding
agent directly from the command line.
💬 Send natural-language requests
📂 Reference files
🛠 Approve tool calls
🔄 Continue multi-step sessions
🧠 Work with configured models
Automation
Headless Mode
Headless mode can run a prompt without
an interactive terminal session. This makes
it useful for scripts, CI/CD and Git hooks.
cn -p "Fix the failing tests"
cn -p "Generate a conventional commit name for the current git changes"
cn --readonly
cn --auto
Continue's current CLI documentation lists
headless prompts, read-only mode, automatic
tool approval, MCP configuration and model
selection flags.
CLI Commands
Useful Continue CLI Options
Command / Flag
Purpose
Use Case
cn
Start interactive mode
Daily development
cn -p "prompt"
Run headless prompt
Automation
--readonly
Read-only tools
Planning and analysis
--auto
Allow tools automatically
Automated workflows
--allow
Allow specific tools
Controlled permissions
--exclude
Exclude specific tools
Restrict capabilities
--mcp
Add MCP server
External tools
--model
Select model
Model customization
--resume
Resume recent session
Continue previous work
--verbose
Verbose output
Debugging CLI workflows
IDE Support
Where Can You Use Continue?
💙 VS Code
Continue provides an IDE extension
for Visual Studio Code.
🧩 JetBrains
Continue has a JetBrains plugin,
although current documentation
recommends the CLI for JetBrains users.
⌨️ Terminal
The Continue CLI provides an
independent terminal coding workflow.
🤖 AI Models
Configure supported models and
providers for different workflows.
🔌 MCP
Connect external MCP servers and
tools to supported agent workflows.
⚙️ Custom Configuration
Customize rules, tools, models and
agent behavior.
Continue's current documentation describes
the VS Code extension, CLI and JetBrains
plugin as its main delivery forms. The
documentation recommends the CLI instead
of the JetBrains plugin.
Models
Flexible AI Model Configuration
Choose Your Model
Continue is designed around configurable
models and providers rather than requiring
one fixed AI model for every task.
🧠 Different coding models
☁️ Cloud providers
🏠 Compatible local models
⚡ Model-specific workflows
💰 Cost-aware model selection
Why Model Choice Matters
Coding quality, context handling, tool use,
speed and cost can vary significantly between
models. Continue's configuration approach
lets developers choose models suitable for
their workflow.
Exact provider and model availability can
change. Check the current Continue documentation
before configuring a particular model.
Permissions
Control Agent Tool Access
❓
Ask First
The default policy can request
permission before tool execution.
⚡
Automatic
Tools can be configured to run
automatically where appropriate.
🚫
Excluded
Tools can be excluded so they
are not provided to the model.
Automatic permissions should be used carefully,
especially for tools that modify files, execute
commands or interact with external systems.
Use Cases
What Can You Do With Continue?
🚀 Build Features
Describe a feature and let Agent
mode work through implementation.
🐛 Fix Bugs
Inspect errors, search the codebase
and implement potential fixes.
🧪 Write Tests
Generate tests and run project
commands to validate changes.
♻️ Refactor Code
Modify multiple files while keeping
the project context available.
📖 Explain Code
Use Chat or Plan mode to understand
unfamiliar source code.
🔎 Analyze Repository
Explore project structure and find
related implementations.
🌐 Research
Use supported web tools to retrieve
useful development information.
🔌 Integrate Tools
Extend workflows with compatible
MCP servers.
⚙️ Automate
Use the CLI in scripts, CI/CD and
Git hooks.
Example Workflow
Fix A Failing Application
The application is failing its tests.
First, inspect the repository and identify
the failing tests.
Do not modify files yet.
Create a plan explaining:
1. What is failing.
2. Which files are involved.
3. Why the failure occurs.
4. What changes should be made.
After the plan is complete, implement
the fix, run the tests again and explain
all changed files.
A useful workflow is to start in Plan mode
for read-only investigation and then switch
to Agent mode when you are ready to make
changes.
Pros & Considerations
Continue Advantages & Limitations
✓ Advantages
Open-source coding agent
Agent, Plan and Chat modes
IDE integration
CLI available
Headless automation
Multiple model configurations
MCP support
Custom rules
Tool permission controls
Codebase search tools
Terminal execution
Useful for CI/CD workflows
! Things To Consider
Results depend on the selected model
Tool support varies by model/provider
Agent mode requires careful permissions
CLI configuration can be technical
Cloud model usage may cost money
Local models require suitable hardware
AI-generated code still requires review
Large repositories can consume significant context
Comparison
Continue vs Other AI Coding Tools
Tool
IDE
CLI
Agent
MCP
Multi-Model
Open Source
Continue
✓✓
✓✓
✓✓
✓✓
✓✓
✓
Cline
✓✓
✓✓
✓✓
✓✓
✓✓
✓
Aider
—
✓✓
✓
✓
✓✓
✓
Claude Code
✓
✓✓
✓✓
✓✓
Primarily Claude
—
Codeium
✓✓
✓
✓
✓
✓
—
Replit
Cloud
✓
✓✓
✓
✓
—
Pricing
Continue Pricing & Cost
Open-Source Software
Continue's coding software is open source.
The cost of using it can therefore depend
on the models and services connected to it.
✓ Open-source project
✓ Custom model configuration
✓ CLI available
✓ IDE integrations
Model Costs
If you connect a paid cloud model provider,
that provider may charge according to its
own pricing and usage rules.
Local models can avoid API charges but
require suitable local hardware.
Always check the current model provider's
pricing before estimating your total AI
coding costs.
Security
Using Continue Safely
🔐
Protect Secrets
Keep API keys, passwords and private
credentials away from prompts and code.
🛑
Review Tool Calls
Inspect commands and file changes before
allowing sensitive operations.
🔎
Inspect Diffs
Review generated modifications before
committing them to the repository.
🧪
Run Tests
Use automated tests to validate
AI-generated changes.
🔌
Audit MCP
Only connect external MCP servers
that you trust.
🌿
Use Git
Keep your project under version control
before making large AI-driven changes.
Best For
Who Should Use Continue?
Great Choice For
💻 Developers who want open-source AI coding
🧠 Users who want model flexibility
⌨️ Developers who like CLI workflows
🔌 Developers using MCP
🛠 Open-source contributors
🚀 Startup developers
🧪 Developers building automated workflows
🏢 Engineering teams
Consider Alternatives If
🎨 You want an AI-first complete IDE
☁️ You want a browser-only development platform
📝 You only need simple autocomplete
🤖 You want a highly autonomous agent with minimal configuration
🔧 You do not want to configure models or tools
Code Smarter With Continue
Use Chat for questions, Plan for safe
repository exploration and Agent mode
for real coding work. Then take the same
agentic workflow into your terminal with
Continue CLI.
Continue is an open-source coding agent
available through IDE integrations and
its command-line interface.
Yes. Continue provides Chat, Plan and
Agent modes for AI-assisted software
development.
Agent mode gives the selected model access
to coding tools that can inspect and modify
the codebase and run commands.
Plan mode provides read-only tools for
safe project exploration and planning
before making changes.
Chat mode is a conversational mode without
coding tools. It can be used for questions,
explanations and discussions.
Yes. Continue provides the cn
command-line coding agent with interactive
and headless modes.
cn is the command used to
launch the Continue CLI coding agent.
Yes. Agent mode includes tools for creating
new files and editing existing files.
Yes. Agent mode can run terminal commands
when the required tool is available and
permission is granted.
Yes. Continue can use MCP servers to
provide additional tools to Agent mode.
Yes. Continue is designed around configurable
models and providers.
Continue supports configurable model
providers, including compatible local
model workflows.
Yes. Continue provides a VS Code extension.
Yes. Continue has a JetBrains plugin.
Current documentation recommends using
the Continue CLI instead of the JetBrains
plugin.
Yes. Continue includes tools such as
glob search, grep search, directory
listing and repository mapping.
Yes. Reading files is one of the
read-only tools available in Plan mode
and is also available to Agent mode.
Yes. The create-new-file tool is available
in Agent mode.
Yes. Because Agent mode can run terminal
commands, it can execute project test
commands when configured appropriately.
Yes. Developers can ask Agent mode to
inspect code, investigate failures and
make appropriate changes.
Yes. Agent mode can modify existing
files and perform multi-step refactoring
tasks.
Yes. Continue CLI supports headless
execution using a prompt flag such as
cn -p "your prompt".
Yes. Headless CLI workflows can be used
for automation, CI/CD and Git hooks.
Yes. By default, Agent mode can ask for
permission before using tools.
Yes. Tool policies can be configured
to allow tools automatically, although
this should be done carefully.
Yes. Tool policies can exclude specific
tools from being provided to the model.
Yes. Continue is an open-source project
released under the Apache 2.0 license.
The Continue software is open source.
However, connected AI model providers
may charge for their usage.
The current CLI documentation supports
Continue account authentication or an
Anthropic API key. Other model-provider
configurations can have their own
authentication requirements.
Yes. MCP allows Continue to connect
additional external tools and services.
Continue can inspect repository information
and diffs, while the CLI can also be used
in Git hooks and automation workflows.
Continue is particularly useful for
developers who want customizable,
open-source AI coding workflows across
their IDE and terminal.
🌐 Connect With TechInfo365
Follow TechInfo365 for technology news,
AI tools, coding guides, blogging,
digital marketing and useful online resources.