Vendo Opens Customization Layer for B2B SaaS
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Vendo debuted as a YC S26 open-source customization layer for B2B SaaS teams whose customers keep asking for bespoke features, letting end users build their own micro-apps without touching the host product's source code.
- The platform runs as an embedded agent that executes through the host product's own API as the signed-in user and renders its generated UI in a sandboxed, brand-native surface.
- Vendo processes work in three stages — Extract (turns the host API into agent tools), Generate (composes views from a format-tagged UI document in an iframe jail with
connect-src 'none'), and Guard (policy, approvals, grants, breakers, and audit at one execution choke point). - Four deployment paths are supported: a tool pack for existing AI SDK/Mastra/homegrown agents, a one-command install for products with no agent, MCP exposure so Claude, ChatGPT, Cursor, and Claude Code act as the signed-in user, and a backend-only package with
agent()andchat()functions. - Vendo stores state in PGlite at
.vendo/datafor zero-config local use and runs the same schema on Postgres in production, with cloud-gated sharing, publishing, org overlays, and pinning activated viaVENDO_API_KEY.
Why it matters: For B2B SaaS teams drowning in bespoke feature requests, Vendo shifts the build burden to end users without exposing source code or breaking brand consistency. The agent runs through the existing API as the authenticated user, with policy guardrails and audit at a single choke point — turning one-off customizations into a self-serve capability rather than a perpetual engineering backlog.
Ask SkimNews
