Vendo

Vendo integrates into an existing application, learns the host product's components, theme, API and permission rules, then exposes those capabilities to an embedded agent. Generated views and micro-apps run through guarded tool access as the signed-in user rather than receiving unrestricted backend privileges. Destructive actions can be approval-gated.

The generated artifact is intended to be persistent software, not a throwaway answer: users can build dashboards, standing automations and integrations with services such as Gmail, Slack and GitHub. Vendo also exposes the host product to external AI clients over MCP so tools such as Claude or ChatGPT can act through the same permissioned surface.

The project is open source and installable through npm. The zero-config development store uses PGlite, while production can use Postgres. Vendo documents sandboxing, policy, approvals, audit, grants and breakers around a shared execution choke point. Enterprise tiers add deeper tenant governance, retention, connectors and identity controls.

Free is $0 forever and includes $5 of monthly usage, roughly 100 typical agent turns, with a hard stop rather than surprise overage. Pro is $49/month with $49 included usage and production publishing/collaboration. Teams is $499/month with $499 usage plus tenant organizations, shared workspaces, roles, approvals, session replay and organizational activity. Enterprise is custom and adds per-tenant controls, SAML/SCIM, DPA/BAA, residency, uptime SLA and deployment on customer infrastructure.

Vendo also publishes unit rates: its cheaper managed models are listed at $1.15/M input and $5.75/M output tokens; default build/edit models at $2.30/M input and $11.50/M output. Sandbox time is $0.01/minute, storage $0.25/GB-month and automation runs $3/1,000. Bring-your-own model/infrastructure paths can avoid corresponding Vendo usage charges.

The reviewed source set produced no substantive current exact-product editorial review, user-review corpus or independent benchmark for Vendo. Product Hunt confirms two 2026 launches and the August 17 relaunch ranking #4 Product of the Day, but currently shows no reviews. That is launch/adoption evidence, not quality validation.

Vendo publishes its own Genbench evaluation of generated product UI: 196 cases across 14 fictional products, comparing Vendo with generative-UI systems and frontier/coding agents. The reported results include 100% data integrity, 99% working screens and the lowest cost per working screen. The benchmark design is useful because it tests generated UI correctness and actionability rather than only language quality, but it is vendor-authored and should not be treated as independent proof.

The open-source repository provides a stronger transparency signal than a closed launch page: implementation details for tool extraction, sandboxing, approvals and storage can be inspected. At capture time the project remains relatively small, so open-source visibility should not be confused with mature ecosystem validation.

Enterprise assurance is still developing. Vendo's pricing page states SOC 2 Type II audit in progress rather than complete. That is an appropriate disclosure, but regulated buyers should treat current compliance maturity as incomplete until the audit is finished.

Vendo is best for B2B SaaS teams whose customers repeatedly request bespoke dashboards, workflows and integrations and who want to expose controlled customization without adding each feature to the core roadmap. It is less suitable for teams that need a proven large-customer track record, mature third-party review corpus or independently reproduced benchmarks before embedding an agent into production.