Continue the conversation — chat opens pre-seeded with the current signal, caps, and movement.
Open-source terminal-first AI coding agent with the fastest community adoption in developer tooling history. Differentiator: provider-agnostic architecture supports 75+ model providers (Claude, OpenAI, Google, local via Ollama) without lock-in. MIT licensed. Built by the SST/Anomaly team (Y Combinator W21). ~178K GitHub stars and ~8M monthly active users (as of June 2026); ~$25M ARR projection disclosed.
OpenCode occupies the open-source alternative position to Claude Code and Cursor. Against Claude Code, it offers model flexibility and open-source transparency but loses on agentic reliability with the same underlying model. Against Cursor, it is terminal-native rather than IDE-embedded. Against Aider, it has superior TUI, desktop app, and team features. The BYOK model makes it the lowest-cost option for teams with existing API keys.
Best fit: Terminal-native developers, teams wanting model provider flexibility, organizations requiring local-first privacy, cost-conscious teams using BYOK. Not ideal for: Teams requiring SOC 2 compliance, organizations needing enterprise SLAs, non-technical users preferring IDE-embedded experiences.
Adoption & Proof Points
- GitHub: ~178K+ stars, 900+ contributors, 13K+ commits, releases every few hours (June 2026). Monthly active users: ~8M (up from 650K Nov 2025, 2.5M Feb 2026, 6.5M Apr 2026, 7.5M May 2026). ARR: ~$25M projection disclosed (June 2026). Fastest developer tool adoption in history per Supabase CEO Paul Copplestone. Enterprise: Cloudflare implementing customized workflows (and pins older versions on several repos per issue #24430), innFactory full-team adoption with internal training program. Backing: Y Combinator W21, undisclosed post-launch funding. Revenue: OpenCode Go ($5 first month then $10/month, now Chinese-lab models only), Zen pay-as-you-go gateway, Black enterprise gateway ($200), enterprise per-seat. SST profitable in 2025. Team: Anomaly Innovations (formerly SST) -- Jay (CEO), Frank Wang (CTO), Dax Raad, Adam Elmore. Controversy: Anthropic blocked OpenCode from Claude consumer OAuth tokens (Jan 9, 2026; ToS update Feb 19; maintainer removed Anthropic OAuth plugin Mar 19), which paradoxically drove awareness; OpenAI partnered to extend Codex subscriptions to OpenCode. Pricing grumbling about the $200 Black tier vs cheaper Codex, but stars/MAU still climbing (no exodus).
Recommended Use Cases
- Terminal-native engineering teams adopting AI-assisted development with existing API keys (BYOK) who want to avoid per-seat model lock-in across Claude, OpenAI, Google, or local Ollama deployments.
- Cost-conscious teams with multi-model strategies: mid-session provider switching and 75+ supported providers allow routing by task type or budget without re-tooling.
- Privacy-sensitive organizations that can accept no compliance certifications — local-first architecture with zero-retention Zen gateway and full MIT source auditability covers the technical bar even where SOC 2 is absent.
- Platform and DevOps teams embedding autonomous CI workflows: GitHub Actions and GitLab CI integrations support autonomous issue triage, feature implementation, PR submission, and line-level code review without a human in the loop.
- Open-source and indie teams building on or extending the agent layer: custom agents via markdown, MCP server support, plugin hooks, and the server question API (v1.4.5) support bespoke integration hosts and community orchestration (OMO).
- Not recommended for regulated enterprises (healthcare, finance, government) requiring SOC 2, HIPAA, RBAC, audit logging, or SCIM, or for non-technical users who depend on IDE-embedded UX rather than terminal-first workflows.
Risks & Limitations
- Enterprise governance gap (Critical): No SOC 2, ISO 27001, FedRAMP, or HIPAA certifications. No RBAC, audit logging, SCIM, or centralized admin. Disqualifies for most regulated environments and enterprise vendor security reviews.
- Security maturity: Two previously patched RCE vulnerabilities indicate security practices still maturing. No bug bounty program documented.
- Agentic reliability: When using the same underlying model, OpenCode loses some agentic reliability vs Claude Code per practitioner comparisons. Auto-compaction regressions (#8234 token consumption) remain open.
- Enterprise readiness: Enterprise tier not yet formally launched. No F500 production case study publicly confirmed. Undisclosed funding amount raises runway questions.
- Anthropic legal risk: Consumer OAuth token blocking (Jan 2026) demonstrates dependency on provider goodwill. BYOK mitigates but does not eliminate risk.
- No persistent memory: No cross-session or cross-repo memory. Context window dependent on selected model.
Capabilities & Integration
Agentic depth: Build mode (full tool access for multi-step implementation) and Plan mode (read-only analysis). General subagent handles complex parallel tasks. GitHub Actions integration executes autonomously from issue/PR comments -- triages issues, implements features, submits PRs, performs line-level code review. GitLab CI integration available. v1.4.0 added --dangerously-skip-permissions for full autonomous mode and OTLP observability. v1.4.2 added clickable subagents. v1.4.5 exposed server question API for custom integration hosts.
Context handling: LSP integration provides type-safe code intelligence (Rust, Swift, Terraform, TypeScript, PyRight). Session management in SQLite. MCP server extensibility. 75+ model providers with mid-session switching. Snapshots now fully respect .gitignore (v1.4.4). OTLP telemetry export for observability (v1.4.5).
Extensibility: Custom agents via markdown files. Plugin system for custom tools and hooks. MCP server support (local and remote). Oh My OpenAgent (OMO) community plugin adds multi-model orchestration with autonomous modes. Enterprise npm registry support (JFrog Artifactory, Nexus).
Release cadence: v1.4.6 as of April 15, 2026. Multiple releases per day during active development. 20+ releases in 33 days (v1.2.17 to v1.4.6).