Kubiya.ai - Open-Source Strategy Deep Dive

1. What Is Open-Source vs. Proprietary?

The Rule: Everything that CALLS the Kubiya API is open. Everything that IS the Kubiya API is closed.

Open-Source (15 public repos under github.com/kubiyabot)

Developer Tools & SDKs:

Infrastructure & Deployment:

Community & Examples:

Proprietary (NOT open-sourced)

• The core platform/control plane -- SaaS backend that orchestrates agents, manages teams, handles auth, runs the AI agent engine
• The AI agent core runtime (ghcr.io/kubiyabot/connections-agent-core) -- private container image
• The web UI/dashboard (app.kubiya.ai / compose.kubiya.ai)
• The agent orchestration engine -- intelligence layer for planning, routing, executing
• Enterprise features -- BYOLLM, advanced governance, SSO, audit trails
• The pricing/billing system and multi-tenant infrastructure

2. Licensing Strategy: Three-Tier Approach

Why AGPL Is Deliberate

AGPL closes the "SaaS loophole" where companies could use GPL code in a cloud service without distributing source. This means:

• Competitors cannot take the CLI/SDK/Terraform provider and build a competing SaaS without open-sourcing their entire service
• Customers can freely use these tools to interact with Kubiya's platform
• Enterprises uncomfortable with AGPL must engage commercially -- AGPL acts as a soft dual-licensing mechanism
• Same strategy used by MongoDB (SSPL), Grafana (AGPL), and Confluent

3. Open-Source Model Classification

Textbook open-core model with AGPL copyleft protection.

• Open periphery: All client-side tools (CLI, SDKs, providers, charts, tools)
• Proprietary core: Control plane, agent engine, orchestration layer
• No "community edition": No free self-hosted version. Open components are useless without the Kubiya API backend. CLI requires KUBIYA_API_KEY, Terraform provider requires KUBIYA_CONTROL_PLANE_API_KEY, MCP server needs a JWT from app.kubiya.ai

Closer to the Databricks/Snowflake model (open client tools, closed platform) than the GitLab model (open core with self-hostable CE).

One exception: skill -- the universal skill runtime (Apache-2.0) can run independently without any Kubiya backend. Functions as a WASM-based tool execution environment for any AI agent. This is a strategic ecosystem play to gain adoption beyond Kubiya's own platform. It has its own website (skill-ai.dev) and works with Claude Code, Aider, Codex, or any terminal-based agent.

4. Strategic Rationale: Why Open-Source the Periphery?

a) Developer Adoption & Ecosystem Lock-in

• CLI, SDK, and Terraform provider let developers integrate using familiar tools
• Terraform modules provide ready-to-deploy solutions that create dependency
• Published to Terraform Registry, PyPI, npm, Homebrew, and Artifact Hub

b) Trust & Transparency

• Enterprise buyers can inspect exactly what CLI/SDK does on their infrastructure
• community-tools shows exactly what tools AI teammates execute -- critical for security-conscious enterprises
• AGPL signals "we're serious about open source" while protecting against free-riders

c) Community-Contributed Tool Ecosystem

• community-tools is the most actively contributed-to repo (15 contributors, 6 forks, 194+ PRs)
• Each contributed tool expands platform capabilities without Kubiya engineering effort
• Tools for Slack, GitHub, JIRA, Kubernetes, AWS, Datadog, etc.

d) Ecosystem Play with skill

• Apache-2.0 licensed, positioned as a universal, agent-agnostic tool runtime
• Strategic bet: if skill becomes standard for how AI agents execute tools, Kubiya benefits as the enterprise orchestration layer on top
• Kubiya's potential "Trojan horse" for broader adoption

e) AGPL as a Commercial Lever

• AGPL is famously problematic for enterprises (legal teams often block it)
• Creates natural demand for commercial licensing
• Same playbook: MongoDB, Grafana, Confluent

5. Community Engagement

Repo Activity

Assessment: Very Low External Community Engagement

• Total stars across all repos: ~32 -- negligible by open-source standards
• Contributors are almost entirely Kubiya employees: shakedaskayo (CTO), mevratavraham, abloch, Kubiya-Barak, raz-alon, Avi-Rosenberg, and others with kubiya in their username
• Zero visible external community contributors outside of Kubiya employees
• No GitHub Discussions activity on any repo
• Open issues mostly bot-generated (dependabot, mend)
• CLI has 30 releases (up to v6.3.10) -- active internal development

Bottom line: Company-driven open source, not community-driven open source. The repos serve Kubiya's customers and sales motion, not an independent community.

6. SDK & Developer Ecosystem

Distribution Channels

Developer Adoption: Low

GitHub engagement metrics (stars, forks, external contributors) all point to minimal external adoption. However, the developer tooling surface is impressively broad -- CLI, Python SDK, Terraform provider, Helm charts, MCP server, npm packages, Homebrew.

7. Comparison to Similar Companies

Kubiya vs. HashiCorp (Terraform)

HashiCorp's BSL move was controversial because Terraform was genuinely community-built. Kubiya's AGPL on client tools is less controversial because the core was never open.

Kubiya vs. Backstage (Spotify)

Kubiya vs. env0

env0 is actually less open than Kubiya. Kubiya's open periphery is a competitive advantage.

Kubiya vs. CrewAI / LangChain

AI agent frameworks are much more genuinely open. Kubiya's approach is enterprise/commercial-first.

Key Takeaways for Your Startup

1. The boundary rule is simple and defensible: client-side = open, server-side = proprietary. Easy to explain to investors, customers, and contributors.

2. AGPL is a strategic weapon, not just a license: It protects against competitors while creating commercial licensing demand from enterprises uncomfortable with copyleft.

3. You don't need community to do open-core: Kubiya has ~32 total stars but uses open source effectively for transparency, developer adoption, and ecosystem integration. Community is a bonus, not a requirement.

4. Consider one Apache-2.0 "Trojan horse": Kubiya's skill runtime is their most strategically interesting bet -- a standalone tool that works without their platform but could drive users toward it.

5. Breadth of developer tooling matters: CLI + SDK + Terraform + Helm + MCP + Homebrew covers every developer persona. This is impressive for a 22-person company.

6. No community edition is a valid choice: Unlike GitLab, Kubiya offers no free self-hosted version. This simplifies the business model and focuses all usage through the paid platform.