# Prized

Build internal tools with AI. Safely.

Source: https://wheredoirunthis.com/products/prized · JSON: https://wheredoirunthis.com/api/products/prized · Read on 2026-08-07

## At a glance

- Type: AI app builder
- Kind: Hosted
- Isolation: Not disclosed
- Free tier: Yes
- Starting price: $100/mo
- GPU: Not documented
- Regions: Not disclosed
- Coverage: 10 of 54 assessed capabilities supported · 7 not assessed
- Agent readiness: read 5 of 11 signals, drive 0 of 11 signals

## Links

- Homepage: https://prized.dev/
- Pricing: https://prized.dev/pricing

## Pricing

- Model: Flat per-workspace subscription with a monthly BUILD allowance, plus optional opt-in pay-as-you-go beyond it. Seats are unlimited and never billed.
- Unit: tool builds / edit sessions per month, per workspace (a fork costs a fraction of a full build; running a deployed tool is free)
- Free tier: Yes

| Tier | Price | Includes | Notes |
| --- | --- | --- | --- |
| Free | $0 | 2 tools a month, no card required; 30 agent messages per tool; unlimited seats; live preview sandboxes; standard connectors (Postgres, Snowflake, Slack, and more).; Seats are unlimited on every tier, so `seats` is null rather than a number. |  [source](https://prized.dev/pricing) |
| Teams | $100/mo per workspace | Everything in Free; 500 tools a month; unlimited agent messages; custom domain for every tool; optional pay-as-you-go extra usage; SSO/SAML; audit log; RBAC and role-scoped connectors; priority support.; Billed per WORKSPACE, not per seat — 'Unlimited seats' on all tiers. |  [source](https://prized.dev/pricing) |
| Enterprise | Custom | Everything in Teams; unlimited usage (fair use); dedicated support and onboarding; invoice billing and custom terms. | 'Custom pricing for organizations that need more than Teams. Contact us.' No public decomposition. [source](https://prized.dev/pricing) |

- Overages: Opt-in pay-as-you-go on paid plans once the monthly build allowance is exhausted; the per-unit rate is disclosed only in the logged-in billing settings, never on the public page. Without opt-in, builds pause and deployed tools keep running.
- Card required: No
- Source: https://prized.dev/pricing

## Runtime

- **GPU & accelerators**: na — "No user-provisioned compute at all — an internal-tools builder exposes no instance sizing, let alone accelerators." (source: product shape) (confidence: high)
- **BYOC (Bring your own cloud)**: No — "No bring-your-own-cloud / customer-VPC deployment mentioned on any public page, including the Enterprise tier ('Custom pricing for organizations that need more than Teams', 'Invoice billing and custom terms' — no deployment-model options listed). Docs gated, so this is a floor." [source](https://prized.dev/pricing) (confidence: medium)
- **Secrets management**: Workspace-scoped connector credentials, approved once by an admin and reused by every tool — no general-purpose secret store documented. — "'Admins approve each connector once and scope what it can see. Every tool you build reuses it.' YC: 'credentials are workspace-scoped so tools can't cross org boundaries.' Nothing about arbitrary user-defined secrets/env vars." [source](https://prized.dev/) (confidence: medium)
- **Storage options**: No first-party data store offered to tools. Tools read/write the customer's connected systems (Postgres, Snowflake, MongoDB, Salesforce, Stripe, Zendesk, Slack, Notion, Google Sheets, HubSpot, Intercom, Asana, Trello, Figma, Dropbox, Microsoft Teams, Zapier). — "17 connector logos on the homepage (/connectors/*.svg|png with 'Move <X> connector' aria-labels); pricing Free tier: 'Standard connectors: Postgres, Snowflake, Slack, and more'. The product's premise is reading EXISTING company data, not provisioning storage." [source](https://prized.dev/) (confidence: high)
- **Supported workloads**: Internal line-of-business tools: record lookups, approval queues, BI-style dashboards, inventory/ops trackers. — "The four worked examples on the homepage — Support ops 'Customer lookup', Finance 'Refund review' (approval queue with risk flags + audit trail), Rev ops 'Billing dashboard' (MRR, churn, revenue per user), Operations 'Stockroom' (stock across three warehouses, reorder-point flags). Not a general-purpose compute target." [source](https://prized.dev/) (confidence: high)
- **Persistence**: Deployed tools are durable and stay live independent of the build allowance; tool DATA is not stored by Prized — tools read the customer's own connected systems of record. — "'Deployed tools stay live'; 'anything you ship stays deployed'; 'deployed tools, their data, and your team's access keep working either way' (when the allowance runs out). Privacy notice: 'We retain workspace content for the life of the workspace and for a short window after deletion to support recovery.'" [source](https://prized.dev/pricing) (confidence: high)
- **Data jurisdiction**: Not documented — "No data-residency, region or jurisdiction statement anywhere on the public surface; the privacy notice names no processing location. Legal entity is a US (Inc.) company HQ'd in San Francisco." [source](https://prized.dev/legal/privacy) (confidence: medium)
- **Resource limits & quotas**: Free: 2 tools/month + 30 agent messages per tool. Teams: 500 tools/month + unlimited agent messages. Enterprise: unlimited (fair use). All plans: unlimited seats. Running a deployed tool is never metered. — "'Every plan includes a monthly build allowance. Starting a new build or an edit session draws on it, and forking an existing tool costs a fraction of a full build. Using a deployed tool never counts. Usage only meters building, and your allowance refreshes in full at the start of each billing cycle.' Unused allowance does not roll over." [source](https://prized.dev/pricing) (confidence: high)
- **SDK languages**: Not documented — "No SDK, client library, npm/PyPI package or programmatic surface named anywhere on the public site. Docs are login-gated (see meta.scan_issues), so this is a floor." [source](https://prized.dev/) (confidence: medium)
- **Network & egress controls**: Outbound calls from a generated tool are proxied through a platform-controlled layer; each tool's reach is limited to admin-approved connectors and an explicit grant list. — "'Each tool gets access only to the tables and integrations it was explicitly granted, credentials are workspace-scoped so tools can't cross org boundaries, and outbound calls go through a controlled layer.' Homepage: 'Tools reach only the production systems an admin has already cleared'; the audit-log demo shows a denial event 'tool.cross_schema_query blocked billing.invoices'." [source](https://www.ycombinator.com/companies/prized) (confidence: high)
- **Snapshots & forks**: Tool-level forking + published versions — not compute-instance snapshots. — "'As an employee builds and shares, a teammate forks it and re-points it at their own data'; pricing FAQ 'forking an existing tool costs a fraction of a full build'; the audit-log demo shows 'tool.published renewal-desk · v3' and the builder shows 'Publish new version'." [source](https://www.ycombinator.com/companies/prized) (confidence: high)
- **Deployment methods**: dashboard — "Chat-in-the-browser only: describe the tool to the agent, watch it build in a live preview, then 'Publish new version'. Controlled vocabulary (features.yml facets.deploy_interface) — no git-push, cli, sdk, api, container-image or template path is advertised." [source](https://prized.dev/) (confidence: high)
- **Isolation**: Not disclosed — "Per the SKILL 'AI app builders' convention this field records the GENERATED TOOL'S HOSTING SUBSTRATE, not a user-selectable isolation boundary — you describe a tool, you do not pick a runtime. The substrate is genuinely undisclosed: the only runtime word on the entire public surface is the Free-tier line 'Live preview sandboxes', which is never defined. (The MARKETING site and app are served by Vercel — `server: Vercel`, `x-vercel-id: fra1::iad1::...` — but that is Prized's own site, NOT evidence about where customer tools run.) Recorded `unknown` rather than inventing a container/VM boundary." [source](https://prized.dev/) (confidence: high)

## Capabilities

### AI & agents

| Capability | Status | Evidence |
| --- | --- | --- |
| Agent skills | Not supported | "No installable/declarable skills or user-defined agent tools; the agent's capabilities are fixed to the admin-approved connector set." [source](https://prized.dev/) |
| AI app builder | Supported | "Describe it. Watch it ship." [source](https://prized.dev/) |
| Autonomous QA agent | Not supported | "Read from Salesforce · Wrote 14 files · Ran 2 checks · 12 steps" [source](https://prized.dev/) |
| Built-in coding agent | Supported | "You describe what you need in plain English, and AI agents build the real thing against your live data." [source](https://prized.dev/blog/backed-by-y-combinator) |
| BYOK for LLMs | Not supported | "No bring-your-own-key option in any tier; agent usage is metered by Prized ('30 agent messages per tool' / 'Unlimited agent messages')." [source](https://prized.dev/pricing) |
| Credential-injection proxy | Not assessed |  |
| LLM gateway | Not supported | "No multi-provider LLM routing/proxy offered; the model powering the build agent is never named." [source](https://prized.dev/) |
| MCP client | Not supported | "The 17 integrations are described as 'connectors', never as MCP servers or MCP tools." [source](https://prized.dev/) |
| MCP-server hosting | Not applicable | "Does not host customer MCP servers as workloads." [source](https://prized.dev/) |
| MCP server | Not supported | "Returns the Next.js sign-in shell (HTTP 200, text/html, 76909 bytes), not an MCP endpoint; the literal phrase 'Model Context Protocol' appears nowhere on the public surface." [source](https://prized.dev/.well-known/mcp) |
| Model inference API (first-party) | Not supported | "Prized consumes a model to drive its build agent; it serves no first-party inference endpoint to customers." [source](https://prized.dev/) |
| Preinstalled coding agent | Not assessed |  |

### Networking & delivery

| Capability | Status | Evidence |
| --- | --- | --- |
| Automatic HTTPS | Partially supported | "Custom domain for every tool" [source](https://prized.dev/pricing) |
| Custom domains | Supported | "Custom domain for every tool" [source](https://prized.dev/pricing) |
| Direct public IPv6 | Not applicable | "No per-instance addressing — tools are reached over managed HTTPS hostnames only." [source](https://prized.dev/) |
| Edge KV store | Not applicable | "Not an edge runtime; no PoP-local KV." [source](https://prized.dev/) |
| Egress controls | Supported | "Each tool gets access only to the tables and integrations it was explicitly granted, credentials are workspace-scoped so tools can't cross org boundaries, and outbound calls go through a controlled layer." [source](https://www.ycombinator.com/companies/prized) |
| Email send / receive | Not supported | "No email send/receive capability named; Slack and Microsoft Teams appear as connectors, email does not." [source](https://prized.dev/) |
| Public preview URLs | Supported | "Live preview sandboxes" [source](https://prized.dev/pricing) |
| Private networking | Not supported | "No VPC, service mesh, private link or service-discovery capability mentioned." [source](https://prized.dev/) |
| Built-in reverse proxy | Supported | "one-click deploy behind your company sign-in" [source](https://prized.dev/) |

### Organisation & security

| Capability | Status | Evidence |
| --- | --- | --- |
| Built-in auth | Supported | "one-click deploy behind your company sign-in" [source](https://prized.dev/) |
| BYOC | Not supported | "No customer-cloud/VPC deployment option in any tier, including Enterprise ('Invoice billing and custom terms' is the extent of the Enterprise differentiation listed)." [source](https://prized.dev/pricing) |
| Deploy into customers' VPCs | Not assessed |  |
| Secrets management | Partially supported | "Admins approve each connector once and scope what it can see. Every tool you build reuses it." [source](https://prized.dev/) |
| Self-host / OSS core | Not supported | "No source repository exists — `gh api orgs/prized` returns 404 and no GitHub repo lists prized.dev as its homepage." [source](https://www.ycombinator.com/companies/prized) |
| Sharing | Supported | "Prized lets non-engineers build, deploy, and share secure internal tools with AI" [source](https://www.ycombinator.com/companies/prized) |
| SSO / SAML | Supported | "SSO/SAML" [source](https://prized.dev/pricing) |
| Teams | Supported | "Every plan includes unlimited seats. Anyone with your company's email domain can join the workspace, covered by the one per-workspace subscription." [source](https://prized.dev/pricing) |
| Workload identity | Not assessed |  |

### Developer access

| Capability | Status | Evidence |
| --- | --- | --- |
| CLI | Not supported | "No CLI named anywhere; the entire interaction model is a browser chat UI." [source](https://prized.dev/) |
| E2B-compatible API | Not assessed |  |
| Filesystem API | Not applicable | "No remote file API; the agent's 'Wrote 14 files' is internal to the build, not a user-facing surface." [source](https://prized.dev/) |
| Git deploy | Not supported | "No GitHub/GitLab connection, repo sync or code export mentioned; the connector list contains no git host." [source](https://prized.dev/) |
| HTTP data API | Not supported | "No HTTP/REST data plane exposed to users; data access happens inside a generated tool via approved connectors." [source](https://prized.dev/) |
| OpenAPI spec | Not supported | "/openapi.json and /api/openapi.json both return the Next.js sign-in shell (HTTP 200, text/html), not a spec." [source](https://prized.dev/openapi.json) |
| Private container registry | Not applicable | "No container/image concept exposed to users." [source](https://prized.dev/) |
| Public API | Not supported | "Disallow: /api/ — the only /api surface is the app's own internal endpoint; no public/programmatic API is documented anywhere on the marketing or pricing pages." [source](https://prized.dev/robots.txt) |
| SDK | Not supported | "No SDK, client library or npm/PyPI package referenced on the public surface." [source](https://prized.dev/) |
| SSH access | Not applicable | "No shell, terminal or instance access — the user is explicitly a non-engineer." [source](https://prized.dev/) |

### Compute & storage

| Capability | Status | Evidence |
| --- | --- | --- |
| Cloud-init / ignition config | Not applicable | "No VM/instance boot configuration surface." [source](https://prized.dev/) |
| Desktop / computer-use sandbox | Not applicable | "No GUI/computer-use sandbox." [source](https://prized.dev/) |
| GPU compute | Not applicable | "No compute sizing or accelerator options anywhere on the public surface." [source](https://prized.dev/) |
| Live-state (memory) snapshots | Not assessed |  |
| Multi-region | Not supported | "No region, data-residency or jurisdiction statement on any public page." [source](https://prized.dev/legal/privacy) |
| Object storage | Not supported | "No S3-compatible bucket or blob store offered to tools; Dropbox and Google Sheets appear only as read CONNECTORS to customer data." [source](https://prized.dev/) |
| Persistent disk | Not applicable | "No user-provisioned compute or attachable volume — you describe a tool, the platform hosts it." [source](https://prized.dev/) |
| Sandbox templates / custom images | Not applicable | "No OCI/Dockerfile base-image concept exposed." [source](https://prized.dev/) |
| Scale to zero | Not supported | "Using a deployed tool never counts." [source](https://prized.dev/pricing) |
| Selectable isolation runtime | Not assessed |  |
| Snapshots / forking | Partially supported | "As an employee builds and shares, a teammate forks it and re-points it at their own data" [source](https://www.ycombinator.com/companies/prized) |

### Data

| Capability | Status | Evidence |
| --- | --- | --- |
| Database branching | Not supported | "Prized provisions no database, so there is nothing to branch; connectors point at the customer's own production systems." [source](https://prized.dev/) |
| Ephemeral instant DB (no-signup) | Not supported | "'Continue with Google' is the only entry point — nothing can be provisioned without an account." [source](https://app.prized.dev/sign-in) |
| First-party database | Not supported | "Company data, pre-connected. Admins approve each connector once and scope what it can see." [source](https://prized.dev/) |
| Vector search | Not supported | "No embeddings, vector index or semantic-search capability named." [source](https://prized.dev/) |

### Platform & operations

| Capability | Status | Evidence |
| --- | --- | --- |
| Managed CI/CD runners | Not applicable | "No CI/CD runner product; no git host is even connected." [source](https://prized.dev/) |
| Built-in observability | Partially supported | "The workspace audit log records who ran each tool, what it touched, and when." [source](https://prized.dev/) |
| Preview environments | Not supported | "No branch/PR model exists — there is a live preview while building and a 'Publish new version' action, but no per-branch ephemeral environment." [source](https://prized.dev/) |
| Scheduled tasks | Not supported | "No cron/schedule primitive named; every worked example is a human-triggered tool ('run.completed renewal-desk · 12 rows')." [source](https://prized.dev/) |
| Project template marketplace | Not supported | "Pick a team to see one built for it. Support ops / Finance / Rev ops / Operations" [source](https://prized.dev/) |
| Durable workflows | Not supported | "Refunds over $150 need review. Give the team an approval queue with risk flags and an audit trail of every decision." [source](https://prized.dev/) |

## Agent readiness

### Read: 5 of 11 signals present

5/11 signals (3 full, 4 partial)

- Llms txt: fail — "GET /llms.txt -> HTTP 200 text/html, the Next.js sign-in shell (76.9 KB), not a text file. No docs subdomain exists to host an alternate copy (rubric v3 allows docs.<domain>/llms.txt — there is no docs host)."
- Llms full txt: fail — "GET /llms-full.txt -> HTTP 200 text/html, same SPA shell (76899 bytes)."
- Robots AI policy: partial — "robots.txt exists: 'User-Agent: * / Allow: / / Disallow: /api/ /admin /sign-in /invite /oauth' + a Sitemap: pointer. Silent on AI user-agents — no GPTBot/ClaudeBot/PerplexityBot/Google-Extended directives either way = default-allow."
- JSON LD: partial — "Homepage ships valid schema.org JSON-LD with an @graph of exactly Organization + WebSite (name, url, logo, description, publisher). No SoftwareApplication/Product/Offer node — the rubric's 'minimal' band."
- Sitemap: partial — "/sitemap.xml is a real XML urlset with 4 URLs (/, /blog/backed-by-y-combinator, /legal/privacy, /legal/terms) and lastmod/changefreq/priority. Valid but <5 URLs = trivial."
- Meta quality: pass — "Complete and descriptive: <title>'Prized - Build internal tools with AI. Safely'</title>, a 226-char meta description, canonical, application-name, theme-color (light+dark), and a full OG set (og:title, og:description, og:site_name, og:image + og:image:type)."
- Markdown negotiation: fail — "Re-requesting / with 'Accept: text/markdown' returns text/html (identical 124674 bytes). /index.md returns the SPA shell, not markdown."
- Content efficiency: partial — "Homepage readable text 3289 chars / 124641 bytes HTML = 2.64% — inside the 1-5% band. Next.js server-rendered, but the markup is animation-heavy and the RSC payload is duplicated inline."
- Docs discoverable: fail — "No docs link in the header, nav or footer (footer is: Guardrails, Audit, Sign in, Schedule a call, Privacy, Terms). /docs 307s to app.prized.dev/sign-in — documentation is entirely login-gated, so an unauthenticated agent can find none of it."
- Pricing machine readable: pass — "/pricing is a real page with three structured, parseable tiers (Free $0 / Teams $100 per mo per workspace / Enterprise Custom), explicit per-tier feature bullets, and 6 FAQ entries defining the usage model. No 'contact us'-only wall."
- Anti bot: pass — "No challenge on any public path — plain curl fetched /, /pricing, /robots.txt, /sitemap.xml, /legal/*, /blog/* at HTTP 200 with no Turnstile/Cloudflare interstitial. Auth surfaces are gated by Google OAuth, which the rubric explicitly does not penalise."

### Drive: 0 of 11 signals present

0/11 signals (0 full, 0 partial)

- Public API: fail — "No public API mentioned on any reachable page. robots.txt disallows /api/ (the console's own internal endpoint). Docs are login-gated, so this is a FLOOR — see meta.scan_issues."
- Openapi spec: fail — "/openapi.json and /api/openapi.json both return the Next.js sign-in shell (HTTP 200, text/html); no spec referenced anywhere."
- MCP server: fail — "/.well-known/mcp returns the SPA shell. No mcp.prized.dev endpoint, no npm/pip MCP package, and the phrase 'Model Context Protocol' appears nowhere on the public surface (rubric v3 guard applied: no bare-substring false positive)."
- SDK packages: fail — "No official or community SDK/client library referenced; no npm or PyPI package named."
- Auth docs: fail — "No API-auth, OAuth-for-third-parties, token or scoped-key documentation is publicly reachable. The only documented auth is human sign-in: 'Continue with Google' (ToS: 'When you sign in with Google we match your account to a workspace by email domain or invitation')."
- Agent self serve creds: fail — "There is no credential for a machine to obtain at all — no API keys, tokens or service accounts exist on the public surface. Human signup IS self-serve and low-friction (Google OAuth, no card, no sales call), but it yields a browser session, not a programmatic credential, so this scores none rather than half."
- Webhooks: fail — "No webhooks, event subscriptions or callbacks mentioned (nor signing docs)."
- CLI tool: fail — "No official or community CLI. The product's entire interaction model is a browser chat UI."
- Error model: fail — "No API documentation exists publicly, therefore no documented JSON error model. Scored fail not na: an internal-tools platform plausibly could expose an API (Retool does)."
- Rate limit docs: fail — "No API rate limits or response headers documented. The only published quotas are commercial build allowances on /pricing (2 vs 500 tools/month), which are not API rate limits."
- Agent config: fail — "/agents.md and /.well-known/agent both return the SPA shell; no skills listing, no agent-facing configuration surface of any kind."

## Compliance & regions

- Regions note: No region, data-residency or jurisdiction information is published anywhere — not on the homepage, pricing page, privacy notice or ToS, and there is no security/trust page (it redirects to sign-in). The marketing site and console are served from Vercel's edge (`x-vercel-id: fra1::iad1::...`), but that describes PRIZED'S OWN site, not where customer tools or connector traffic are processed, and must not be read as a region disclosure. Notable gap given the product connects to production systems of record for finance and support teams.

## Company & presence

- Headquarters: San Francisco, CA, USA
- Team size: 2 (both founders; no other team members disclosed)
- Privacy policy: https://prized.dev/legal/privacy
- Terms of service: https://prized.dev/legal/terms
- Marketing site: https://prized.dev/
- Pricing page: https://prized.dev/pricing
- Blog: https://prized.dev/blog
- Y Combinator company page: https://www.ycombinator.com/companies/prized

## Nearby in the atlas

- [AppDeploy](/products/appdeploy.md)
- [Base44](/products/base44.md)
- [Lovable](/products/lovable.md)
- [Playcode](/products/playcode.md)
- [Remy](/products/goremy.md)

## Method

Facts come from the product's own public pages. Each claim carries its evidence: a quote and the page it came from. We count capabilities rather than grade products, and "Not assessed" means we have not checked, never that the product lacks it.