DataFog
DataFog builds open-source tools that keep sensitive data out of AI workflows. Its products detect PII and credentials locally, then redact, mask, remove, pseudonymize, tokenize, or otherwise enforce privacy policy before data reaches an agent, model, log, API, or another system.
At a Glance
- AI-agent and LLM application developers
- Developer-tool and CLI builders
- Security and compliance teams needing local/runtime DLP
- Organizations preparing or sharing customer data and datasets
- +1 more
AI Tools by DataFog
(1)
DataFog Python
Python PII Redaction Library
Discussions
No discussions yet
Be the first to start a discussion about DataFog
Latest News
DataFog MCP presented as now available for Claude Code
DataFog Core 0.4.4 released with aligned Rust, Python, Node, and WASM runtime tags
DataFog Core 0.4.3 released
DataFog Core 0.4.2 released with published multi-runtime packages
Products & Services
Rust-based shared privacy engine for detecting PII in text and structured records and applying redaction, masking, removal, keyed pseudonymization, provider-backed tokenization, and restoration. Distributed with Rust, Python, Node.js, and browser/WASM SDKs.
Local MCP server for Claude Code that checks sensitive files before an agent opens them and creates redacted, masked, or cleaned copies without returning file contents or matched values to the agent.
Python library and offline PII firewall. It provides regex-based structured detection, optional spaCy and GLiNER NER, agent-oriented scan/filter APIs, a Claude Code hook, and a LiteLLM gateway guardrail.
Offline Claude Code firewall plugin that uses the DataFog hook to protect agent tool calls from emails, credit cards, SSNs, and other sensitive entities.
Market Position
DataFog positions itself as a local, developer-first privacy engine and data-governance layer for AI workflows: it runs in-process or locally, avoids a required sidecar/service, and combines detection with configurable enforcement and transformations. Its Python repository explicitly benchmarks against Presidio and spaCy NER; compared with those detector-oriented alternatives, DataFog emphasizes agent hooks, gateway guardrails, process-boundary policy enforcement, and a shared Rust engine across SDKs.
Leadership
Founders
Sid Mohan
Founder and CEO of DataFog. His public profiles identify San Francisco as his location and list UCLA Anderson School of Management; public search results also associate him with prior work at Oakhurst Ventures, On Deck, and One Drop.
Executive Team
Sid Mohan
Founder and CEO
Public profiles identify Mohan as DataFog's CEO/founder, based in San Francisco, with UCLA Anderson School of Management listed in his LinkedIn profile; he is also the primary public maintainer of the DataFog repositories.
Founding Story
DataFog was started to provide a privacy tool that scans data for sensitive information and redacts or replaces it with synthetic data. The initial vision was local, developer-oriented protection for AI and LLM workflows rather than a network-dependent service.
Business Model
Revenue Model
The public offering is open-source software distributed through GitHub and package ecosystems under permissive licenses. The public materials emphasize self-hosted/local execution and do not document a paid subscription or usage-based pricing model.
Pricing Tiers
DataFog Core, DataFog MCP, DataFog Python, and several integrations are publicly available; Core and MCP repositories state MIT licensing, while the Claude Code plugin states Apache-2.0.
Target Markets
- AI-agent and LLM application developers
- Developer-tool and CLI builders
- Security and compliance teams needing local/runtime DLP
- Organizations preparing or sharing customer data and datasets
- Teams using Claude Code, LiteLLM, OpenClaw, Rust, Python, Node.js, or browser applications
- Protecting prompts, tool calls, trace payloads, and model outputs in AI-agent workflows
- Preparing customer exports, datasets, and structured records for analysis or sharing
- Checking outbound API requests, uploads, and exports before PII leaves an environment
- Guarding shell commands and developer-tool process boundaries from credential or PII leakage
- Runtime DLP and programmable privacy enforcement for security and compliance teams
- Local file scanning and protected-copy creation through Claude Code