Continue.dev
Continue is an open-source autopilot tool for software development that can be used as a Visual Studio Code (VS Code) extension. It integra...
Last verified:
What is Continue.dev?
Continue.dev provides quality control for software development by running source-controlled AI checks on every pull request. Engineering teams define standards as simple markdown files in the .continue/checks/ directory within their repository, each containing a name, description, and specific prompt instructing the AI on what to inspect. When a PR is opened, Continue automatically executes these checks against the code diff, posting results as native GitHub status checks—green for passing code or red with actionable suggested fixes if issues are detected.
This approach ensures consistency by only enforcing predefined rules without unsolicited feedback, scaling effortlessly as development velocity increases. Teams focus on high-level design and architecture rather than routine reviews, as mechanical enforcement is handled automatically. Suggested fixes appear directly in GitHub PRs for one-click acceptance or rejection, streamlining the review process.
Continue.dev targets engineering teams and software factories aiming to maintain code quality at scale. It is ideal for organizations with growing codebases where manual reviews become bottlenecks, enabling humans to set AI-enforced standards. Developers benefit from immediate feedback on common issues like security vulnerabilities or style violations without disrupting workflow.
Continue.dev pricing
Pricing model: Free
Free to use with quickstart at continue.dev/check; no paid plans or tiers mentioned on the website—open source with automatic GitHub integration.
Continue.dev pros
- Source-controlled checks in repo
- Native GitHub status integration
- One-click fix suggestions
- Scales with development speed
- Enforces only defined standards
- No unsolicited AI opinions
- Markdown-based check definition
- Automatic PR diff analysis
- Reduces review queue backlog
- Focuses team on architecture
- Prevents surprise bugs consistently
- Customizable prompts per check
- Security scanning for secrets
- Input validation enforcement
- SQL injection detection
- Logging sensitivity checks
- Easy quickstart at continue.dev/check
Continue.dev cons
- Limited to GitHub PRs currently
- Requires .continue/checks/ setup
- Dependent on AI model accuracy
- No support for other VCS yet
- Manual check writing needed
- Potential false positives in checks
- Fix suggestions may need review
- Scales based on prompt quality
- GitHub-only status checks
Frequently asked questions about Continue.dev
What is Continue.dev?
Continue.dev runs AI-powered checks on every pull request, using markdown files in .continue/checks/ to define standards. Each check appears as a GitHub status—green if code passes, red with suggested fixes if not, enabling automated quality control decided by humans.
How do you set up checks?
Create markdown files in .continue/checks/ with YAML frontmatter for name and description, followed by a prompt detailing what to flag, like hardcoded secrets or missing validation. Continue runs these automatically on PRs.
What happens on a failed check?
A red GitHub status check posts with details on the issue and a suggested code fix. Users can accept or reject the fix directly in the PR interface without manual intervention.
Who is Continue for?
It's for software factories and engineering teams needing scalable quality control, where checks enforce standards automatically so developers focus on design instead of routine reviews.
Does it integrate with GitHub?
Yes, checks post as native GitHub status checks on PRs, fully integrated for visibility and interaction within the GitHub workflow.
Can checks catch security issues?
Yes, example checks flag hardcoded API keys, tokens, passwords, new endpoints without input validation, string-concatenated SQL queries, and sensitive data in logs.
Is there a quickstart?
Go to continue.dev/check to test checks on a pull request immediately, then add markdown files to your repo for ongoing automation.
What makes it consistent?
Checks only enforce what the prompt specifies, never missing defined issues and avoiding generic or unsolicited AI feedback on unrelated matters.
How does it scale?
Checks adapt automatically to increasing PR volume and faster factory speeds, maintaining quality without falling behind or requiring extra resources.
Are checks human-decided?
Yes, standards are written by humans as markdown prompts; AI merely enforces them reliably without adding opinions.