Univer

```json
{
  "service_type": "platform",
  "base_url": "https://univer.ai",
  "auth_method": "unknown",
  "auth_config": {
    "note": "No public authentication documentation found in provided content. Requires investigation of docs or platform access.",
    "contact": "developer@univer.ai"
  },
  "endpoints": [],
  "pricing_model": {
    "type": "unknown",
    "details": {
      "note": "Open-source core available on GitHub (dream-num/univer). Commercial 'Univer Platform' offering likely exists but pricing not disclosed on provided page.",
      "open_source": "https://github.com/dream-num/univer"
    }
  },
  "rate_limits": {},
  "capabilities": [
    "AI-native office runtime for agents and humans",
    "Spreadsheets (Excel-class: 500+ functions, PivotTables, 10M cells, 29 chart types)",
    "Documents (Word-class layout, DOCX import/export, Notion-like blocks, LaTeX)",
    "Slides (PowerPoint-class fidelity, PPTX import/export)",
    "Canvas/Boards (Miro-style, flowcharts, UML, mind maps, 200+ shapes)",
    "Relational Table (Airtable-style: Grid/Kanban/Calendar/Gantt/Gallery views, 22 field types)",
    "PDF viewing, editing, forms, and export",
    "Real-time collaboration between humans and multiple agents",
    "Multi-agent isolated worktrees with review, diff, merge, rollback",
    "Univer as Code: declarative, structured, versioned file repository",
    "Cross-unit data references with live synchronization",
    "XLSX/DOCX/PPTX import and export",
    "Server-side and browser rendering",
    "CLI skills integration (npx skills add dream-num/skills)"
  ],
  "raw_analysis": "Univer (univer.ai) is an ambitious open-source-rooted platform positioning itself as an 'office harness for AI agents' — an agent-native runtime unifying spreadsheets, docs, slides, canvases, relational tables, and PDFs into a single `.univer` file format. It targets developers building AI agent workflows and teams wanting human-agent collaboration, explicitly contrasting itself with fragmented Office suites (Excel/Word/PowerPoint), Miro, Airtable, and Notion by claiming to consolidate those workflows into one runtime with agents as first-class citizens.\n\nMaturity assessment: The core spreadsheet engine is mature (10M cell performance, 500+ functions, Excel parity claims, active GitHub project under dream-num/univer with a large community). The newer agent-native framing (multi-agent worktrees, Univer as Code, unified runtime) appears to be a strategic pivot/expansion and likely earlier-stage — marketing language is aspirational. Company: DreamNum (Chinese/US developer-tools organization).\n\nAPI surface: No public REST API endpoints are documented in the provided material. Integration paths appear to be (1) the open-source JS/TS SDK, (2) the `skills` CLI package (dream-num/skills) which suggests MCP or agent-tool integrations rather than REST, and (3) direct file-level operations on `.univer` files. Auth model is undocumented here.\n\nPricing: Not disclosed. Open-source core is freely available; commercial 'Univer Platform' presumably offers hosted/enterprise tiers. Contact developer@univer.ai.\n\nRate limits: Unknown.\n\nRecommended next steps for Prowl: (1) check GitHub repo for SDK/API surface, (2) inspect package.json of dream-num/skills to understand agent-integration mechanism, (3) request API docs / pricing from developer@univer.ai, (4) probe for a hosted platform endpoint at app.univer.ai or similar."
}
```

3/3 tests passed

```json
{
  "overall": 78,
  "dimensions": {
    "token_efficiency": 8.5,
    "first_try_success": 7.5,
    "response_parseability": 9.0,
    "error_clarity": 7.5,
    "doc_quality": 8.5,
    "auth_simplicity": 7.0,
    "latency": 9.5,
    "consistency": 8.0
  },
  "pricing_normalized": {
    "model": "open_source_core_plus_undisclosed_commercial",
    "open_source": true,
    "pricing_visibility": "low",
    "note": "Open-source core (dream-num/univer on GitHub) is verifiable; commercial platform pricing not disclosed. Agents can't quote costs — a friction point for recommendations."
  },
  "issues": [
    "No public pricing for the commercial 'Univer Platform' — hard to give users cost expectations.",
    "No auth/API-onboarding details surfaced (llms.txt covers capabilities, not signup or API keys).",
    "Missing security headers on responses (no HSTS/CSP observed), which slightly lowers trust for agent-driven integrations.",
    "The value prop is very broad (sheets+docs+slides+tables+canvas+PDF) — risk of cognitive overload when describing concisely.",
    "No explicit rate limits, quotas, or failure-mode docs visible in the fetch."
  ],
  "recommendations": [
    "Publish a pricing page with tiered/free/commercial breakdown to make it recommendable by agents with cost context.",
    "Add API-key/magic-link auth docs to llms.txt so agents can onboard programmatically without human hand-holding.",
    "Add a compact 'capabilities → format → endpoint' cheat sheet to llms.txt for downstream tool-use.",
    "Document rate limits, quotas, and error codes; this raises first_try_success significantly.",
    "Enable standard security headers (HSTS, CSP, X-Content-Type-Options) to build trust.",
    "Add a one-sentence canonical pitch at the top of the marketing site (e.g., 'Notion + Excel + Slides, with native multi-agent worktrees') to cut token cost when recommending."
  ]
}
```