One governed MCP for all your data sources.
Most teams give AI tools data access by adding one MCP server per tool, each wired straight to the data with its own rules, or none at all. Rig replaces that sprawl with a single governed layer, so every person and every agent gets only the access their role allows.
Why one governed MCP
Connecting MCPs directly
- A separate MCP server per tool and per source
- Each one connects straight to raw data
- No shared access model. Governance is per-tool, or missing entirely
- Source credentials end up in the agent's context
- No unified record of what was read, by whom
Rig's governed MCP
- One endpoint for Claude, Cursor, ChatGPT and other AI tools
- Every person and agent authenticates first, with OAuth or scoped, revocable tokens
- Access resolved by role on every call
- PII and financial fields masked, row filters applied at query time
- Credentials stay inside Rig, never in model context
- Every call written to an audit trail
How every request is governed
The same four steps run on every query, whether it comes from a person or their agent. Nothing runs unless access is explicitly allowed.
Authenticate
Every connection, whether a person or an agent, is identified with OAuth or a scoped, revocable token. Admins can switch off any user's agent access without touching their login.
Resolve role
Rig looks up what that identity is allowed to see, across schema, table, column and row. Explicit deny always wins.
Rewrite & enforce
The query is rewritten before it runs: blocked tables rejected, PII and financial columns masked, row filters injected. Read-only by default, with writes limited to edit roles. If access is ever unclear, the query is denied.
Audit
The call, the SQL, the tables touched and the masks applied are all logged, alongside every change to permissions.
The controls behind it
Administrative, technical, and organisational controls appropriate to the risk, the same ones described in our Product Privacy Policy.
Encryption
Encrypted in transit with TLS 1.2 or higher, and at rest with AES-256 (or provider-equivalent) on Rig-controlled infrastructure.
Role-based access
Roles govern access at schema, table, column and row level. New data defaults to least-privilege, and explicit deny beats any allow.
PII & financial masking
Columns are classified by sensitivity and masked per role, by nulling, hashing, partially showing or redacting the value. Masking is enforced when the query runs.
Credential isolation
Source keys live in a managed secrets store (AWS Secrets Manager) and are never passed through the MCP or into model context.
Audit trail
Every data access and every permission change is logged with the actor, what was touched, and what was masked. The log is queryable for review.
Incident response
A documented incident-response process, with breach notification within the period set out in the applicable agreement.
Where your data lives
Two deployment models, depending on your setup. You decide what Rig can access, and where it runs.
Connect your own warehouse
Your data stays in your environment. Rig processes it transiently, such as query results and representative sample values, to deliver the service, without persistently replicating or hosting your warehouse.
Rig-managed ingestion & hosting
Ask us to ingest and host your data and it lives in an isolated per-tenant warehouse managed by our partners at MotherDuck, under the controls above.
In both models, processing is limited to what's necessary, and your data is deleted or returned on termination.
Sub-processors
The third-party sub-processors we engage to provide the Service. Each is bound by data-protection obligations equivalent to ours and receives only what it needs to perform its function.
| Sub-processor | Purpose | Region |
|---|---|---|
| MotherDuck | Isolated per-tenant data warehouses for Rig-hosted customers | EU / US |
| Amazon Web Services (AWS) | Context cache and secrets management | EU / US |
| Supabase | User accounts and conversation cache | EU |
| Langfuse | LLM traces and reasoning logs | EU |
| Sentry | Error monitoring | EU |
| PostHog | Product analytics (team emails and user IDs) | EU |
| Nango | Authentication and API calls for permitted integrations | US |
One governed door to your data.
See how governed, AI-native access works on your own warehouse and tools.