Memory
CattleGrid (cattlegrid.uk) — competitor validation and strategy for DocuClear. Found via the Qwen deep-research report 'Beyond ChatGPT: A 2026 Guide to Cost-Effective Sovereign AI in the UK and EU' (which miscategorised CattleGrid as a voice/speech specialist — the report's gateway section misses security gateways entirely; early, poorly-mapped market). What CattleGrid is: UK-registered company (Cheltenham, Company No. 17020793, no overseas parent) selling an AI API security gateway — network-level interception of team AI traffic (OpenAI, Anthropic, Gemini, Mistral), memory-only scanning with pass/block/redact/flag actions, sub-8ms target, zero retention, UK-only processing (Civo UK, Frankfurt option), GDPR pseudonymisation framing, volume pricing (free launch tier, from £50, no per-seat, on-prem on custom tier). They submitted written evidence to a UK Parliament AI inquiry. Why it validates DocuClear: same 'say yes to AI safely' thesis as DocuClear's 'make ChatGPT safe to use' positioning; same regulatory anchors (UK GDPR, DPA 2018, ICO principles, EU AI Act); same audience — small UK orgs without compliance teams. Strategy decision: DocuClear + Cloudflare AI Gateway + XBerg redaction = CattleGrid competitor. Cloudflare natively provides the gateway half (AI Gateway Guardrails + DLP, Cloudflare One AI prompt topic detection for PII/credentials/financial/source code, WAF NER-based PII detection, prompt logging by conversation_id). XBerg redaction covers the redact action — redact-at-source in the workspace beats CattleGrid's redact-in-transit. DocuClear adds what CattleGrid lacks: persistent workspace where flagged work gets resolved, grounded ICO/EU AI Act retrieval via MCP, watermarked PDF audit trails, agent collaboration. Killer feature idea: reversible pseudonymisation — XBerg redacts at source, token map stored in workspace vault, output re-identified locally; auditable GDPR Art 4(5) pseudonymisation a zero-retention proxy structurally cannot offer. Caveats: CattleGrid's sharpest wedge is UK jurisdiction vs Cloudflare being US-incorporated (CLOUD Act) — need Cloudflare data-localisation story and careful gateway log config; must match free/£50 pricing expectations. Next steps: (1) route DocuClear LLM traffic through AI Gateway with guardrails/DLP for a demo-able 'CattleGrid moment' inside the workspace; (2) expose XBerg redaction as an MCP tool (redact_document with token-map write-back); (3) positioning: 'CattleGrid stops the leak; DocuClear does the work safely and proves it.' [Update Sep 3, 2026]: Update (Sep 3, 2026, ~6:10am BST): CATEGORY CONTEXT via Perplexity — the 'agent/AI gateway governance' category is forming FAST in 2026, which validates the thesis but shortens the window: Okta launched Agent Gateway (runtime agent governance: brokered credentials, attributed tool calls, revocation) in July 2026; Microsoft open-sourced an Agent Governance Toolkit (MIT, runtime policy enforcement, OWASP agentic-risk coverage) in April 2026; Databricks has Unity AI Gateway as 'enforcement fabric'; Gartner is publishing on agent harnesses as the governance layer. READ: the demand signal is real and being ratified by majors — but 'gateway that watches AI traffic' will commoditise quickly under Okta/Microsoft/Cloudflare pressure. DocuClear's defensible position is therefore NOT the gateway (CattleGrid's whole product; everyone else's feature) but the governed WORKSPACE above it: where flagged work gets resolved, where work happens with provenance, where redaction is reversible via the workspace vault (the killer feature noted below), and where SMB utility (docs/site/phone/media/automations) makes the safety layer a by-product rather than the purchase. CattleGrid validates the problem; the wedge in entry 330253d7 (usefulness first, safety as accelerant) is the answer to a commoditising gateway market.
Tags: docuclear, cattlegrid, competitor-analysis, cloudflare, xberg, strategy, volatility:state, project:docuclear, kind:semantic — Source: mcp — 2026-09-02 19:17:20 UTC