Redactly
An API that automatically strips personal information like names and addresses out of text before it's sent to an AI model, so sensitive data never actually reaches the AI provider.
🔗 Visit Redactly
Description
When a company connects its data to an AI model like ChatGPT or Claude, there's a real risk that private details — a customer's name, address, or account number — end up sitting in a prompt sent to a third party. Redactly sits in between: it scans the text first, replaces the sensitive parts with safe placeholders, and only then lets the sanitized version go through to the AI, so the AI (and whoever stores its logs) never actually sees the real personal data.
Redactly is a privacy middleware API for detecting and masking PII in text before it reaches LLMs, vector databases, logs, or analytics pipelines, with a single API call sitting in front of providers like OpenAI or Anthropic. It performs detection and redaction locally rather than relying on an external LLM or third-party service to do the scanning, preserves context using descriptive labels or optional synthetic substitution so prompts remain useful after redaction, and supports bi-directional redaction on both requests and model responses. It's positioned to help engineering teams support GDPR, HIPAA, and SOC 2 compliance programs.
💬 Our review
The short version: Redactly targets a real and growing problem — companies wiring their own data into LLMs without a clear plan for keeping personal information out of those calls — with a simple, single-API-call fix.
Doing detection locally rather than routing text through another external LLM to find PII is a sensible design choice, since it avoids just moving the privacy problem to a second third party. Preserving context with placeholder substitution instead of just deleting matches is also the right call for keeping the redacted prompt actually usable by the model. The honest gap is transparency: there's no public pricing page, the homepage blocked automated research tools during our check, and as an early-stage product it doesn't yet have the track record of a company like Portkey, which bundles PII redaction as one guardrail among many in a more established platform. If you specifically need a lightweight, dedicated PII-redaction layer and are comfortable reaching out directly for pricing, it's worth evaluating; if you want an established, feature-rich AI gateway with redaction as one piece, Portkey is the safer bet today.
📊 Global score
🤖 AI-enriched data
Aucune page de tarifs publique trouvée ; contact direct nécessaire pour connaître le prix.
Pros
Détection et masquage effectués localement, sans passer par un LLM tiers
Substitution contextuelle qui garde le prompt exploitable après redaction
Fonctionne à la fois sur les entrées et les sorties du modèle
Pensé pour soutenir la conformité GDPR/HIPAA/SOC 2
Cons
Aucune tarification publique disponible
Site qui bloque les outils de récupération automatisée, évaluation en libre-service difficile
Produit jeune, peu de recul comparé à des plateformes établies comme Portkey