Kadoa

Kadoa is the most AI-native company in this segment, and the only one that sells collection rather than data. Its Assistant, launched in June 2026, turns a plain-language request into a tested pipeline that repairs itself when a website changes, and every value links back to its source page. An MCP server shipped in February 2026 and has been extended ten times since, and Enterprise clients receive data in Snowflake, S3 or Databricks. A Compliance Center, automated robots.txt checks and SOC 2 Type 2 address the approval step that slows most web-data projects.

What Kadoa does not supply matters as much under this framework. There is no proprietary panel and no back history: a dataset starts on the day its pipeline is built. We did not locate ticker mapping, MNPI or personal-data screening, or any predictive-value evidence, and every performance figure is company-reported for unnamed clients. Funding is seed-stage, and the round led by Vi Partners in March 2025 was undisclosed.

We initiate at 6.38, Emerging, with an Improving outlook. Product Capability (7.00) and Workflow Fit (6.83) lead its profile, with AI & Technology at 6.42; Data & Content, at 5.67, is held back by the variant's history and ticker-mapping criteria. Named references and documented entity mapping would move the score most.

Strengths

• Agentic collection. The Assistant explores a site, picks the most reliable source, proposes a schema, builds and tests the pipeline and waits for approval before going live.
• Self-healing pipelines. Generated code is regenerated when sources change, with escalation to a person only when recovery fails.
• Value-level lineage. Deterministic code rather than per-page model calls, with each value traceable to its source and edit history kept.
• AI-assistant delivery. An OAuth MCP server usable from Claude, ChatGPT, Cursor and Codex, extended almost monthly since February 2026.
• Compliance workflow. A Compliance Center for reporting and rule monitoring, automated robots.txt checks and approval workflows.
• Easy to trial. A free trial and a no-commitment evaluation period on every plan.

Cautions

• No history. Datasets begin when a pipeline is built, so they cannot be back-tested until they have run for some time.
• Mapping. We did not locate ticker mapping or entity resolution; buyers should expect to map output to securities themselves.
• Screening. MNPI and personal-data screening are not documented; scraping terms-of-use risk remains with the client's compliance review.
• Evidence. Clients are described as top-five hedge funds and market makers but none is named, and outcome figures are company-reported.
• Model disclosure. The model providers behind the Assistant are not named.
• Scale. Seed-stage, with funding amounts and headcount undisclosed.
• Assurance detail. SOC 2 reports and policies are available only on request; on-premise deployment appears only in a blog post.