Documentation
The full documentation site lives behind the customer console (authenticated, with embedded API reference) and ships publicly at general availability. Everything you need for the design-partner pilot is here.
Sections
Quickstart
Deploy the OCI image to your container runtime. Provide a Postgres or MySQL connection string, an API key for the control plane, and your customer alias. The proxy reports ready inside 90 seconds and pulls policy on its first /sync.
See pricing tiers →Connectors
Open the console, register a source, give it an alias, and paste the credentials. The proxy validates the connection and surfaces sample schema. Add a contract or skip and let the built-in PII fallback masking apply until you do.
Contracts
Open Data Contract Standard v3 is the canonical format. Tag PII columns (pii.email, pii.phone, phi.diagnosis), set freshness SLAs, define per-role field policies. The console validates the contract against the live source schema before saving.
Blueprints
Eight blueprints ship today: HIPAA, SOC 2, FINRA, GDPR with UK GDPR, CCPA with CPRA, DPDP Act 2023, LGPD, and PIPEDA. Each one sets trust thresholds, mask defaults, and audit retention to framework-aligned values. Apply once. Override per source where you need stricter (or looser) defaults than the blueprint prescribes.
SDKs
Python and TypeScript SDKs wrap the proxy with idiomatic clients. MIT licensed, on GitHub. The SDK manages keys, retries, and the trust-aware lookup endpoint so your agent code stays clean.
Integrations
Copy-paste MCP server configs for Claude Desktop, Cursor, Continue, and Cline. Function-calling tool definitions for the Anthropic Messages API, OpenAI Chat Completions, OpenAI Responses, LangChain, and the Vercel AI SDK.
See integration snippets →Reference
OpenAPI 3.1 reference, auto-generated from the @datadam/shared Zod schemas. Operation summaries, request and response shapes, and auth requirements for every customer-facing endpoint.
Open API reference →LLM egress DLP
The proxy fronts agent traffic to Anthropic, OpenAI, and Gemini. Every outbound prompt is scanned for PII, secrets, and operator-authored patterns. Image attachments route through an in-proxy detection pipeline: extracted text gets the same scan as prompts, and detected regions are painted over before forwarding. Per-vendor mode (warn / redact / block), the CPU or GPU deployment choice, and the org-level image toggle all live on the Settings page.
Read the full capability page →Upgrade safety
The procurement-friendly doc on what state survives a version bump, what the rollout looks like, what we promise not to change without warning. Forwardable to your platform-engineering lead with no editorializing.
Read upgrade safety →Deployment topology
Per-region, per-environment, per-business-unit, and failover topologies. The single proxy comfortably serves hundreds of millions of requests per month; multi-proxy is usually a posture question, not a throughput question.
Read deployment topology →Email hello@mydatadam.com for the technical brief, the security architecture deck, and the design-partner SOW.