TokenVeil's roadmap
TokenVeil moves fast. Here is the road travelled since June, release by release, and where we're heading next.
v0.0Shipped
Birth: the anonymized multi-AI chat
- Self-hosted chat with reversible anonymization across 8 providers: Claude, Gemini, OpenAI, Mistral, GitHub Models, Azure OpenAI, Vertex AI, Bedrock.
- Drag-and-drop attachments, screenshot paste, structured document preview.
v0.1Shipped
The security and compliance toolkit
- VS Code / Cursor extension, with API keys (expiry, daily quota, revocation).
- Public, reproducible anonymization benchmark.
- GDPR compliance pack for DPOs and CISOs.
- Standalone reversible file anonymization, OCR included (PNG, JPEG, scanned PDF).
- P0 hardening: rate limiting, strict headers, vendored assets (air-gap ready), per-IP login lockout.
v0.2Shipped
Ready for enterprise sale
- Dynamic model catalog: new provider models appear on their own, retired ones disappear.
- Commercial licensing: Ed25519 license gate, bytecode-only Docker delivery.
v0.3Shipped
Structured detection and speed
- Key-driven detection on structured data: in JSON and key-value logs, a sensitive key (name, phone, iban) is enough to tokenize the value.
- De-anonymization rewritten: streaming on log-heavy conversations went from ~7 s to ~0.12 s.
v0.4Shipped
Anonymization becomes an API
- Public API v1: anonymize and translate back from your own pipelines, with encrypted, per-key, opaque sessions.
- Translate module: anonymize here, work in your own tool (Claude Code, ChatGPT), translate the reply back. Expiring, revocable, audited share code.
- License hardening: three bypasses closed after an internal audit.
Todayv0.4.0-alpha · 5 releases since June
In progress
What a first large deployment expects
- Independent security audit, then fixing the findings.
- Audit log export to your SIEM (Splunk, Sentinel, syslog) and a monthly compliance report for the DPO.
- SSO SAML / OIDC (Okta, Microsoft Entra) and SCIM provisioning.
- PostgreSQL and Kubernetes deployment (Helm chart) for scale.
- Encryption key managed through a KMS / Vault.
Planned
Protecting AI wherever people touch it
- Browser extension: protection directly inside ChatGPT, Claude, and Gemini on the web.
- OpenAI-compatible gateway: your internal applications protected without a rewrite.
- Anthropic-compatible gateway for coding agents (Claude Code, Cursor): anonymization at the API level, the agent itself is untouched.
- MCP server: anonymization as a tool for agentic workflows.
- Self-hosted LLMs (Ollama, vLLM) so even the model stays home.
- Real-time compliance dashboard with alerts.
- Per-role policies and industry dictionaries (healthcare, banking, legal).
- Slack and Teams connectors.
Vision
What becomes hard to copy
- In-house fine-tuned NER engine, trained on French corpora and business data.
- A small, fully local LLM to catch ambiguous PII, with nothing ever leaving.
- Extended multilingual support (German, Spanish, Italian).
- Secure RAG: anonymization upstream of embeddings and vector stores.
- SOC 2 and ISO 27001 certifications.
Indicative roadmap, prioritized with our first partners: the first feature built after a pilot is the one that pilot requires. A direction, not a promise of dates.
Try TokenVeil in real conditions
Leave your email, we'll reach out directly to talk about your use case. No generic pricing shown here, we discuss it together.
contact@tokenveil.eu
Share this with your team: download the one-pager (PDF)