Castari

What do they actually do

Castari provides a managed platform to deploy and run AI agents in sandboxed environments with guardrails and controlled access to external tools/skills (including MCP-based tools). Their YC listing positions them as “Vercel for AI Agents – Deploy sandboxed agents with MCP tools & skills,” and the public site says they help teams “deploy production ready AI agents with predictable guardrails” (YC company page, castari.com, Anthropic on MCP).

Publicly, access is via a contact email rather than a self-serve signup or docs, suggesting the product is in private/early access at this stage (castari.com).

Who are their target customer(s)

• Early-stage product teams adding an agent feature to their app: They need to ship an agent to production without building custom infra and worry the agent might call the wrong APIs or leak user data.
• Mid-size SaaS engineering/platform teams: They want a consistent, auditable runtime so multiple teams don’t create ad-hoc, fragile deployments and access patterns.
• Internal automation teams at regulated companies (finance, legal, HR): They must enforce strict access controls, logging, and approvals to meet compliance and fear unsafe agent actions in sensitive systems.
• Developers who build connectors and integrations: They want a predictable runtime to expose tools safely to agents and avoid re-implementing one-off integrations for each project.
• Security/ops teams responsible for production reliability: They need kill-switches, rate limits, and observability because model-driven automations can behave unpredictably and are hard to monitor.

How would they acquire their first 10, 50, and 100 customers

• First 10: Founder-led, hands-on pilots via YC/network: embed an engineer to deploy one agent with a couple tools, de-risk with bespoke guardrails/NDAs, and convert pilots into referenceable case studies and referrals.
• First 50: Invite-only beta with repeatable 4–6 week paid pilots; targeted outreach (LinkedIn/GitHub/communities/YC alumni), plus starter templates (Slack/CRM/DB) and 2–3 connector partners that recommend Castari; standardize a pilot-to-paid checklist.
• First 100: Light self-serve (signup, templates, docs) and a paid pilot SKU; hire an SDR and a CS engineer to run demos/onboarding; run focused webinars/workshops by vertical and amplify wins into case studies and event presence to close larger platform/regulated accounts.

What is the rough total addressable market

Top-down context:

Castari sits between AI developer tooling and enterprise automation. As a proxy, the AI development tool software market is projected in the low-to-mid tens of billions over the next few years, while RPA/automation remains large and is adding “agentic automation” offerings (Statista, UiPath agentic platform announcement).

Bottom-up calculation:

Using ~$13B as a conservative proxy for AI dev tools, a managed agent-runtime slice at 1–10% implies ~$130M (1%), ~$650M (5%), and ~$1.3B (10%) TAM scenarios for platforms like Castari (Statista).

Assumptions:

• Managed agent runtimes capture roughly 1–10% of AI dev tools spend in the medium term.
• Castari addresses the runtime/guardrails/observability slice, not model training or raw cloud compute.
• Buyers split budgets across devtools and automation; in-house builds and cloud incumbents limit near-term penetration.

Who are some of their notable competitors

• LangGraph Cloud (LangChain): Hosted/hybrid/self-hosted deployment options for LangGraph agents with observability and enterprise controls; LangChain also announced a LangGraph Cloud offering (LangGraph page, LangGraph Cloud announcement).
• Dify: Open-source agentic workflow builder with a managed cloud, templates, and a marketplace for tools/connectors to develop, deploy, and operate agents (Dify homepage, Docs).
• Amazon Bedrock Agents: AWS service to build and configure agents that securely connect to company data, call tools/APIs, and run multi-step workflows with guardrails (Bedrock pricing page referencing Agents).
• Azure AI Foundry Agent Service: Azure’s agent service for creating and hosting AI agents with tools, multi-agent orchestration, and enterprise security/compliance (Microsoft product page).
• UiPath (Agentic Automation): Enterprise automation vendor extending into agentic automation for large organizations; relevant for buyers consolidating agent runtime and orchestration with RPA (announcement).