Council vs Spider Cloud

The two products barely overlap in capability, which is the whole story. Spider Cloud's value is the rendering and retrieval layer: a scrape can return markdown, JSON, HTML, raw, or plain text, and return_format markdown strips nav, ads, and footers. Browser mode runs scripts, waits on lazy images, and follows infinite scroll to completion, while protected pages route through a real browser engine with geo checks. Crawls stream each page as one JSONL line the moment it finishes rather than buffering, and data_connectors pipes results into S3, GCS, Google Sheets, Azure Blob, or Supabase as pages land. AI extraction is two-phase — a fast model for most pages, a stronger model for complex layouts — with extraction_schema forcing output into your JSON schema. Newer additions include a provider router (Sept 2026) that can hand URLs to outside providers on your own keys, Browser AI commands (Act, Extract, Observe) over the WebSocket, and a catalog of 1,000+ ready-made scraper examples. Council's value is epistemic, not infrastructural: twelve backends answer independently, then critique each other blind, producing a 0–100 divergence score showing how many camps formed and who the outlier is. It adds a devil's advocate role, a neutral chair that writes the synthesis instead of answering, a guest seat for pasting outside answers, vision and .md/.txt attachment, and a local decision journal that later records how the call turned out. One is plumbing for web data; the other is a deliberation ritual for humans.

Spider Cloud is freemium with usage-based tiers, and its distinguishing pitch is flat-rate Unlimited concurrency versus metered per-request billing — attractive for high-throughput crawls where per-request pricing punishes volume. Caveats live in the plan detail, not the marketing: geo-targeting isn't available on the Unlimited plan yet, Unlimited queues nothing server-side so your app must handle HTTP 429 itself, and Browser AI commands require an AI Studio subscription on top. There's also a real cost of ownership in the follow-the-rules default — Spider follows robots.txt unless you disable it per request. Council inverts this: the app is free, and your spend is entirely your own LLM API keys or a local runtime. That makes it cheap if you already pay for Claude, GPT, Gemini, DeepSeek, Grok, Mistral, Perplexity, or OpenRouter, and free at the margin if you run Ollama or a self-hosted OpenAI-compatible endpoint — the privacy-conscious setup is also the zero-marginal-cost one. The honest comparison is that Spider Cloud's bill scales with pages crawled, while Council's scales with how often you ask the same question of several models. Neither replaces the other's line item.