Steedos Platform

{
  "service_type": "platform",
  "base_url": "https://www.objectstack.ai",
  "auth_method": "unknown",
  "auth_config": {
    "notes": "Docs reference OAuth and API key for MCP client setup; environment routing suggests per-environment auth contexts. Exact auth scheme not fully specified in provided content."
  },
  "endpoints": [],
  "pricing_model": {
    "type": "unknown",
    "details": {
      "notes": "No pricing information present in provided content. Platform appears to be an enterprise B2B product (Steedos / ObjectStack)."
    }
  },
  "rate_limits": {},
  "capabilities": [
    "Metadata-driven enterprise application development",
    "Visual no-code app building (Console, Studio, Setup) with code-sync of metadata",
    "ObjectStack layer: ObjectQL, ObjectOS, Object UI",
    "REST data API (CRUD, batch, clone, analytics queries) over every object",
    "GraphQL APIs generated per object",
    "Metadata API (object schemas, UI views, package management)",
    "TypeScript Client SDK with auto-discovery and typed metadata",
    "Declarative custom REST endpoints as metadata",
    "Environment routing (dev, preview, prod)",
    "Plugin architecture with per-plugin REST endpoints",
    "Automation: flows (DAG), hooks, workflows, approvals, scheduled jobs, durable webhooks",
    "Approval chains with multi-step sign-off and audit trail",
    "Connectors (rest, openapi, mcp) for external systems",
    "AI agents (ask and build) with skills, MCP tools, and HITL approvals",
    "RAG via Knowledge Protocol with pluggable adapters (memory, RAGFlow, custom)",
    "MCP client support for Claude Code, Claude Desktop, Cursor, IDEs",
    "Natural language queries over live data with row-level security",
    "Analytics: ~20 chart families, dashboards, reports, TV display pages",
    "Views: table, kanban, calendar, gantt, gallery, timeline, map, tree, chart",
    "Permissions: 4-layer model, per-field visibility, RLS, sharing rules, explain API",
    "Forms with validation, wizard layouts, public forms",
    "40+ field types with auto-computed fields and relationships",
    "Email templates and notifications",
    "Import/export, webhooks, federated external databases",
    "Marketplace templates",
    "i18n and internationalization",
    "Realtime events and notifications",
    "Package management and metadata lifecycle (load, publish, HMR)",
    "Upgrade paths between major versions",
    "Error catalog with structured error envelopes"
  ],
  "raw_analysis": "Steedos Platform (branded here as ObjectStack) is an AI-native infrastructure for enterprise applications. The product positioning — 'Turn Prompts into Enterprise Software' — places it in the emerging low-code/no-code + AI agent platform category, competing with the likes of Retool, Airtable, Budibase, Appsmith, and increasingly AI-first app builders like Vercel v0 and Cursor-driven stacks, though the metadata-protocol framing is closer to Salesforce Platform / ServiceNow / Microsoft Power Platform in ambition.\n\nCore architecture is organized around four protocol layers (Data, System, UI, and presumably AI/Automation) with a metadata-driven runtime: developers and business analysts author structured metadata (objects, fields, views, flows, permissions, endpoints, agents, skills, tools) either visually or as code, and the runtime generates APIs (REST + GraphQL), UI, automation, and AI-agent tools from that single source of truth. This is a protocol-first, ontology-as-code design — the docs repeatedly frame it as 'one executable business ontology — AI writes it, the runtime runs it, agents operate it, you own it.'\n\nTarget audience is clearly enterprise: the docs cover RLS, HITL approvals, approval chains, audit trails, environment routing (dev/preview/prod), package management, upgrade guides between major versions, and a full error catalog. That maturity signals a platform designed for governed production deployments rather than hobbyist app building.\n\nAPI surface is substantial. Beyond generated per-object REST and GraphQL, there is a Metadata API for publishing packages, a declarative endpoint system for custom routes, plugin endpoints (auth, workflow, automation, views, realtime, notifications, AI, i18n, files), a TypeScript Client SDK with auto-discovery, and MCP servers exposing actions as AI-callable tools. Authentication is not fully documented in the provided content — the MCP client setup page references OAuth or API key, and environment routing implies scoped credentials, but the exact scheme (token vs session vs OAuth flow) is not confirmed here.\n\nAI is treated as a first-class primitive rather than a bolt-on: two built-in platform agents (ask, build), structured Skills as reusable instruction modules, declarative Action metadata exposed as MCP tools with explicit opt-in and HITL approval, permission-aware RAG via a pluggable Knowledge Protocol, and natural-language query tools (query_records, get_record, aggregate_records) that run under

3/3 tests passed

{
  "overall": 78,
  "dimensions": {
    "token_efficiency": 7.0,
    "first_try_success": 6.5,
    "response_parseability": 8.0,
    "error_clarity": 7.0,
    "doc_quality": 8.0,
    "auth_simplicity": 5.0,
    "latency": 9.0,
    "consistency": 7.5
  },
  "pricing_normalized": {
    "pricing_model": "unknown",
    "note": "No pricing metrics present in provided data; enterprise B2B positioning makes cost opaqueness a friction point for agent recommendations."
  },
  "issues": [
    "Pricing model unclear — no published tiers or cost signal, making budget-aware recommendations difficult.",
    "Auth onboarding path not documented in provided content; likely complex enterprise setup given B2B positioning.",
    "Capability surface is very broad (metadata platform + flows + MCP + RAG + analytics), which risks overwhelming a user in one description pass."
  ],
  "recommendations": [
    "Publish a concise 'what is ObjectStack in one sentence' hero block to improve token efficiency for AI agents.",
    "Add a public pricing page or free-tier signup to boost first_try_success and agent recommendability.",
    "Expose magic-link or notlogin.com-style onboarding to reduce auth friction.",
    "Surface a status page and uptime metrics to strengthen consistency perception.",
    "Keep llms.txt / llms-full.txt updated (already a strength) and add a machine-readable capability index for structured retrieval."
  ]
}