Continue the conversation — chat opens pre-seeded with the current signal, caps, and movement.
Cline fork that evolved into a differentiated multi-mode agentic IDE. Launched as "Roo Cline" in late 2024, rebranded to Roo Code in early 2025. Positions as "a whole dev team of AI agents" versus Cline's single-agent paradigm. 1.1M+ VS Code installs, 21.3K GitHub stars, 283 contributors. Series A startup (~$6M funding, ~9 employees) based in San Rafael, CA. CEO Matt Rubens.
Adoption & Proof Points
- 1.1M+ VS Code Marketplace installs
- 21.3K GitHub stars (vs Cline's 56K)
- 6,454 total commits (47% more than Cline's 4,376)
- 283 contributors
- ~15K Discord members
- 50+ releases in past 3 months (v3.30-v3.42)
- Works in VS Code, Cursor, Windsurf, VSCodium
- No named enterprise customers documented
- No published case studies
Recommended Use Cases
- Teams wanting Cline-like experience with more customization
- Organizations requiring local-first/BYOK architecture
- Developers wanting multi-model flexibility with per-mode assignment
- Workflows benefiting from autonomous Cloud Agents (PR review, CI fixes)
- Teams comfortable with early-stage vendor risk
Risks & Limitations
- Company maturity: ~9 employees, $6M funding—small compared to competitors
- Security posture: 7 CVEs in 2025, SOC 2 Type II incomplete
- Enterprise gaps: No ISO 27001, HIPAA, or FedRAMP. SAML only on Enterprise tier.
- Adoption lag: Trails Cline in installs and stars despite technical advantages
- No enterprise validation: Zero named customers or case studies found
- Dependency risk: Small team maintaining rapidly evolving codebase
Capabilities & Integration
Multi-mode architecture: Five specialized modes—Code (implementation), Architect (system design), Ask (Q&A), Debug (troubleshooting), Orchestrator (workflow coordination). Each mode has distinct tool permissions and system prompts. Mode-based context isolation prevents pollution.
Boomerang Tasks: Orchestrator mode enables autonomous workflow orchestration. Analyzes complex tasks, breaks into subtasks in isolated contexts, coordinates results. Prevents "context poisoning" in long-running tasks. Each subtask returns only summaries to parent.
Cloud Agents: 24/7 autonomous agents in isolated containers. Five types: Coder (generates code, creates PRs), PR Reviewer (automatic code review), PR Fixer (resolves comments/CI failures), Planner (implementation plans), Explainer (documentation). GitHub App integration for full PR workflow. Slack integration for task initiation.
Context handling: Semantic codebase indexing with configurable embedding providers. Concurrent 100-file reads (v3.17+). Smart Code Folding (v3.40) preserves function signatures during context condensation. @-mentions for files, folders, git commits, URLs.
Model support: 30+ providers including Anthropic, OpenAI, Google, AWS Bedrock, DeepSeek, Mistral, xAI. Full BYOK. Local models via Ollama, LM Studio, OpenAI-compatible endpoints. API Configuration Profiles assign different models per mode.
MCP support: Full implementation with global and project-level config. STDIO and HTTP transports. Custom MCP server creation supported.
